EU AI Act Enforcement Timeline: Key Deadlines from February 2025 to August 2028
The EU AI Act enters phased enforcement with prohibited practices taking effect February 2025, followed by GPAI governance rules in August 2025 and high-risk system regulations by December 2027.
The EU AI Act’s Phased Entry into Force
The EU AI Act entered into force on 1 August 2024, establishing the regulatory framework for artificial intelligence across the bloc. However, enforcement takes place in stages, with different provisions applying on different dates.
February 2025: Prohibited Practices and AI Literacy Obligations
Prohibited AI practices and AI literacy obligations will enter into application from 2 February 2025. This marks the first enforcement milestone, addressing immediate harms and establishing baseline literacy requirements across organisations.
Notably, AI systems that generate non-consensual sexually explicit and intimate content or child sexual abuse material are prohibited under the Act.
August 2025: GPAI Governance Rules
The governance rules and the obligations for GPAI (general-purpose AI) models will become applicable on 2 August 2025. The European AI Office’s powers will be reinforced at this stage, with the office gaining centralised oversight of AI systems built on general-purpose AI models.
December 2027: High-Risk AI Systems
Rules for systems used in certain high-risk areas will apply from 2 December 2027. These high-risk applications include biometrics, critical infrastructure, education, employment, migration, asylum, and border control.
August 2028: Integrated AI Systems
For systems integrated into products such as lifts or toys, the rules will apply from 2 August 2028.
Support for SMEs and Innovation
Simplified technical documentation requirements for SMEs and small mid-cap companies are extended under the regulatory framework. Additionally, more innovators will gain access to regulatory sandboxes, including an EU-level sandbox, to support compliance and innovation.
May 2026: AI Act Amendments
A political agreement on the AI Act amendments was reached on 7 May 2026, known as the Digital Omnibus on AI, reflecting evolving policy priorities across the EU Parliament, Council, and European Commission.
Ireland’s Implementation Framework
Ireland will adopt a distributed model of competent authorities for the AI Act, as approved by Government Decisions of 4 March 2025 and 22 July 2025.
A new statutory independent body called the AI Office of Ireland will be established under the Department of Enterprise, Tourism and Employment. The AI Office of Ireland will serve as the Single Point of Contact and central coordinating authority for the implementation and enforcement of the EU AI Act in Ireland.
Developments since publication
-
The AI Office published a list of over 180 organisations that signed the Code of Practice on transparency of AI-generated content. Source
-
In January 2026, the European Commission opened a formal investigation into X under the Digital Services Act over its Grok tool. Source
-
The EU AI Act entered into force on 1 August 2024 and became applicable on 2 August 2026. Source
-
EU AI Act prohibitions on harmful AI practices (items 1–8, including social scoring and real-time remote biometric identification) entered into application from 2 February 2025. Source
-
On 30 July 2026, the EU launched an AI Gigafactories call to boost Europe's computing capacity and unlock more than €30 billion in investment. Source
-
The Commission's July 2026 action plan on Cybersecurity and AI includes a planned call to increase EU evaluation capacity of AI models before they are placed on the EU market, expected to be operation Source
-
The European Commission and ENISA will create a blueprint to secure access to advanced AI systems for cybersecurity purposes and establish a secure testing platform for organisations in critical secto Source
-
The AI Omnibus extends certain simplified compliance requirements — previously available only to SMEs — to small mid-cap companies, including simplified technical documentation requirements. Source
-
The AI Omnibus centralises oversight of AI systems built on GPAI models under the AI Office, reducing governance fragmentation. Source
-
Article 50 of the EU AI Act began applying on 2 August 2026, requiring AI systems that interact directly with people to disclose that users are dealing with AI unless the context makes that obvious. Source
-
Certain law enforcement activities are exempt from some of the transparency obligations introduced by Article 50 of the EU AI Act. Source
-
In May 2026, EU lawmakers agreed to postpone the high-risk AI obligations — covering biometrics, employment, education, essential services, and migration, asylum and border management — until 2 Decemb Source
-
High-risk AI systems in the delayed categories remain subject to existing laws such as GDPR and sector-specific regulation, but will not yet have to comply with the AI Act's dedicated high-risk govern Source
-
EU Executive Vice President Henna Virkkunen stated the aim of the high-risk delay was to 'make it easier to innovate without lowering the bar on safety', arguing companies and regulators needed cleare Source
-
The European Commission linked the Digital Omnibus delay to Mario Draghi's 2024 report on European competitiveness, which argued that the EU regulatory burden was holding back growth across the econom Source
-
European Parliament negotiators backed the Digital Omnibus compromise on the grounds that the technical standards underpinning compliance for high-risk AI systems were not ready in time. Source
-
Annex III of the EU AI Act classifies certain AI systems used in migration, asylum, and border management as 'high risk', covering tools used for assessing risks, assisting decisions on asylum, visa a Source
-
Obligations for general-purpose AI (GPAI) models under the EU AI Act began applying in August 2025. Source
-
China's Implementation Opinions on the Standardized Application and Innovative Development of Intelligent Agents became enforceable on July 15, 2026. Source
-
China's Implementation Opinions on Intelligent Agents were issued jointly by the CAC, NDRC, and MIIT. Source
-
China's Implementation Opinions on Intelligent Agents is widely described as the first national policy document to treat AI agents as a distinct regulated category. Source
-
Under Article 6 of China's Implementation Opinions on Intelligent Agents, an agent's decision authority must be sorted before deployment into three tiers: decisions only a human may make, decisions th Source
-
Agents in sensitive sectors such as healthcare, transport, media, and public safety face additional filing, testing, and recall requirements under China's Implementation Opinions on Intelligent Agents Source
-
New York's Department of Financial Services issued Insurance Circular Letter No. 7 (2024), requiring insurers to run and document a comprehensive assessment before using AI or external consumer data i Source
-
In 2026, New York's DFS layered on further guidance addressing cybersecurity and frontier-AI risks for AI used in regulated financial decisions. Source
-
There is no comprehensive federal AI statute in the United States as of mid-2026. Source
-
India's Digital India Act, intended to replace the Information Technology Act of 2000 with a risk-based framework covering AI, algorithmic transparency, and deepfakes, remained in draft and consultati Source
-
The EU has officially transitioned into the enforcement phase of its AI Act, targeting high-risk applications such as facial recognition and predictive algorithms. Source
-
Bipartisan disagreements in the United States prevent the establishment of a cohesive federal AI framework, leaving states and local jurisdictions to fill the regulatory void. Source
-
The European Commission's AI Office, together with national authorities, began enforcing the AI Act on 2 August 2026. Source
-
Deepfakes — images, videos, or audio edited or generated using AI — are now required to be labelled under the AI Act transparency rules. Source
-
The Code of Practice on transparency of AI-generated content is voluntary, but the transparency obligations it operationalises are legal requirements under Article 50 of the AI Act. Source
-
Enforcement responsibility under the AI Act is split: the AI Office handles GPAI models directly; national competent authorities handle other AI systems operating within their borders; the European Da Source
-
All providers of GPAI models must document certain information and provide it to competent authorities or downstream providers, and must put in place a copyright policy. Source
-
National competent authorities are responsible for enforcing AI Act rules on AI systems operating within their own borders, distinct from the AI Office's GPAI mandate. Source
-
The Digital Omnibus on AI entered into force on July 27, 2026. Source
-
The Digital Omnibus on AI was published in the Official Journal as Regulation (EU) 2026/1744 on July 24, 2026. Source
-
Article 50 transparency obligations became enforceable on August 2, 2026. Source
-
The EU AI Office's enforcement powers over general-purpose AI (GPAI) models became active on August 2, 2026. Source
-
From August 2, 2026, the EU AI Office can request information, demand model access, and impose penalties on GPAI model providers. Source
-
Fines for transparency and GPAI violations reach up to 15 million euros or 3% of total worldwide annual turnover, whichever is higher. Source
-
AI systems that interact directly with people must make clear that a person is dealing with a machine, unless that is obvious from context. Source
-
Deepfakes and manipulated audio, image, or video content must be disclosed as artificially generated under Article 50. Source
-
AI systems already on the EU market before August 2, 2026 have a grace period until December 2, 2026 to comply with the machine-readable content-marking requirement. Source
-
High-risk obligations for stand-alone Annex III AI systems (including recruitment screening, credit scoring, education, and law enforcement tools) now apply from December 2, 2027. Source
-
High-risk obligations for AI embedded in regulated products (Annex I) now apply from August 2, 2028. Source
-
A new prohibition on AI systems generating non-consensual intimate imagery was added by the Digital Omnibus and takes full effect from December 2, 2026. Source
-
The prohibited AI practices ban has applied since February 2, 2025. Source
-
Obligations for general-purpose AI model providers have applied since August 2, 2025, but the Commission could not act on violations until August 2, 2026. Source
-
Breaches of the AI Act's new transparency obligations can attract administrative fines of up to 15 million euros ($17.3m) or 3 percent of global annual turnover, whichever is higher. Source
-
AI systems used to recognise emotions or categorise people using biometric data must inform individuals that such processing is taking place. Source
-
The AI Act's high-risk obligations for AI systems used in biometrics, employment, education, essential services, and migration were postponed in May 2026 as part of the Digital Omnibus package. Source
-
The AI Act's high-risk obligations for systems in areas such as biometrics, employment, education, and migration were delayed until December 2, 2027. Source
-
Annex III of the EU AI Act classifies AI systems used in migration, asylum and border management as 'high risk'. Source
-
Prohibited AI practices and rules on AI literacy under the AI Act began applying in February 2025. Source
-
Obligations for general-purpose AI models under the AI Act took effect in August 2025. Source
-
From 2 August 2026, chatbots and other interactive AI systems are required to tell users they are dealing with AI, not a human. Source
-
From 2 August 2026, deepfakes — images, videos, or audio edited or generated using AI — must be labelled. Source
-
From 2 August 2026, AI-generated or altered content must carry machine-readable marks so it can be detected more easily. Source
-
Companies that ignore Article 50 transparency obligations risk fines of up to €15 million or 3% of their worldwide annual turnover, whichever is higher. Source
-
Henna Virkkunen, Executive Vice-President for Tech Sovereignty, Security and Democracy, warned that the most advanced AI models 'create risks on an entirely new scale.' Source
-
The AI Office's enforcement powers cover providers of general-purpose AI (GPAI) models and include the power to request technical documentation, run evaluations, demand corrective steps, and issue fin Source
-
Providers of the most advanced GPAI models must address risks of large-scale harm, including chemical, biological, radiological and nuclear incidents, loss of control, cyber offence, harmful manipulat Source
-
All providers of GPAI models must document certain information and provide it to competent authorities or downstream providers, put in place a copyright policy, and publish a sufficiently detailed sum Source
-
The Code of Practice on transparency of AI-generated content is voluntary, though the transparency obligations under Article 50 are legal requirements regardless of whether a provider signs it. Source
-
The AI Omnibus pushed back rules for high-risk AI systems to 2 December 2027. Source
-
The AI Omnibus pushed back rules for high-risk AI systems built into regulated products to 2 August 2028. Source
-
From 2 December 2026, the AI Omnibus bans AI systems that generate non-consensual sexually explicit content or child sexual abuse material. Source
-
In January 2026, the European Commission opened a formal investigation into X under the Digital Services Act over its Grok tool, after manipulated sexually explicit images and possible child sexual ab Source
-
The European Data Protection Supervisor is responsible for overseeing EU AI Act compliance among EU institutions themselves. Source
-
Henna Virkkunen said the AI Act gives 'innovators legal certainty while protecting the public interest' and called enforcement 'an important step towards AI that people and businesses can understand a Source
-
New transparency rules under the EU AI Act started to apply on 2 August 2026. Source
-
The AI Office holds enforcement powers over GPAI models, including the ability to request technical documentation, evaluate models, require corrective measures, and issue fines for non-compliance. Source
-
A July 2026 action plan on Cybersecurity and AI sets out a coordinated approach to help Member States, businesses and public authorities address cybersecurity and resilience challenges posed by the mo Source
-
The Digital Omnibus on AI deferred full compliance for high-risk AI systems listed in Annex III — including recruitment tools, credit scoring, education, law enforcement, and border control systems — Source
-
AI embedded in products already covered by EU product safety law under Annex I — such as medical devices, machinery, and toys — faces a compliance deadline of 2 August 2028. Source
-
Implementation of the high-risk AI regime stalled because member states were slow to designate national competent authorities, and the harmonised standards and conformity assessment tools that high-ri Source
-
The Article 50 chatbot disclosure requirement — requiring organisations to inform users they are interacting with an AI — applies from 2 August 2026 to any organisation deploying a chatbot or AI assis Source
-
Improper handling of personal data by AI systems, especially in biometric or emotion recognition applications, may lead to GDPR-related penalties of up to EUR 20,000,000 or 4% of annual worldwide turn Source
-
From 2 August 2026, the European Commission's AI Office, together with national authorities, will begin enforcing the Artificial Intelligence (AI) Act. Source
-
From 2 August 2026, new transparency rules require certain AI systems to tell users when they are interacting with AI and when content has been generated or altered by it. Source
-
Chatbots and other interactive AI systems will have to tell users they are dealing with AI, not a human. Source
-
Deepfakes (images, videos, or audio that have been edited or generated using AI) will have to be labelled under the new AI Act transparency rules. Source
-
AI-generated or altered content will have to carry machine-readable marks so it can be detected more easily. Source
-
The Commission published a first list of more than 180 organisations that have signed the Code of Practice on transparency of AI-generated content. Source
-
The Commission's supervision and enforcement powers against GPAI model providers came into force on 2 August 2026, following a one-year adjustment period during which GPAI model obligations were alrea Source
-
Providers of GPAI models released before 2 August 2025 must be compliant before 2 August 2027. Source
-
The Commission has exclusive powers to supervise and enforce obligations under Chapter V of the AI Act pursuant to Article 88 of the AI Act. Source
-
The maximum fine that may be imposed on GPAI model providers under Article 101 of the AI Act is 3% of their annual total worldwide turnover in the preceding financial year, or €15,000,000, whichever i Source
-
A GPAI model is presumed to have high impact capabilities if the cumulative training compute of the model exceeds 10^25 FLOP. Source
-
Providers of GPAI models with high impact capabilities must notify the Commission without delay and in any event within two weeks after that requirement is met or it becomes known that it will be met. Source
-
Providers of GPAI models released under a free and open-source licence are only required to comply with the copyright policy and training content summary obligations, unless the GPAI model presents a Source
-
Providers of GPAI models with systemic risk (GPAISR) must also perform model evaluations, conduct risk assessment and mitigation, record and report serious incidents, and ensure a sufficient level of Source
-
The AI Office's scientific panel may issue a qualified alert requiring at least a simple majority of its members, and after receiving such an alert the AI Office must decide whether to exercise its en Source
-
For providers of GPAI models that adhere to a code of practice assessed as adequate, the Commission stated it will focus its enforcement activities on monitoring their adherence to the code of practic Source
-
Providers established in third countries are required to appoint an authorised representative in the EU prior to placing their GPAI model on the EU market, unless the model is released under a free an Source
-
Under the Digital Markets Act decision, search data sharing begins in January 2027, with Android interoperability due by July 2027. Source
-
Prior Labs was founded by Frank Hutter, Noah Hollmann, and Sauraj Gambhir roughly 18 months ago. Source
-
Prior Labs' TabPFN model series was published in Nature and set the state of the art on tabular benchmarks across hundreds of independent academic studies. Source
-
Oracle signed a $300 billion five-year cloud contract with OpenAI covering 4.5 gigawatts of capacity as part of Stargate, the $500 billion AI infrastructure initiative. Source
-
Gemini 3.5 Pro reportedly missed its July 17 target, marking the third slip for Google's flagship model. Source
-
Moonshot AI's Kimi K3 has a 2.8-trillion-parameter open model that took the top spot on a major coding leaderboard. Source
-
Moonshot AI has promised Kimi K3's open weights by July 27, 2026. Source
-
Epoch AI tested three leading AI text detectors—Pangram, GPTZero, and Originality.ai—against text generated in imitation of a specific author's style and found that up to 18 percent of AI-generated pa Source
-
A new benchmark called RadLE 2.0 found that AI models frequently deliver wrong findings in radiology tasks with full confidence, reading X-rays and producing incorrect diagnoses without any signal of Source
-
Prohibited AI practices under the AI Act became effective in February 2025. Source
-
AI systems that generate non-consensual sexually explicit and intimate content or child sexual abuse material have been prohibited following amendments agreed in the AI Omnibus. Source
-
Chatbot disclosure requirements became live and binding on July 10, 2026, requiring users to be clearly informed they are interacting with an AI system at the start of any interaction. Source
-
Violations of prohibited AI practices under the EU AI Act carry maximum fines of €35 million or 7% of global annual turnover, whichever is higher. Source
-
Violations of high-risk AI obligations under the EU AI Act carry maximum fines of €15 million or 3% of global annual turnover, whichever is higher. Source
-
The Digital Omnibus confirmed that AI systems using personal data must comply with both the AI Act and GDPR, with GDPR supervisory authorities and AI Act national competent authorities expected to coo Source
-
The European Commission presented the EU Action Plan on Cybersecurity and Artificial Intelligence on July 7, 2026, to strengthen the enforcement of the AI Act by establishing dedicated evaluation capa Source
-
The evaluation capacity from the EU Action Plan is expected to become operational in 2027. Source
-
Enforcement of key provisions for advanced models begins in August 2026, requiring developers to demonstrate risk evaluations and mitigation strategies. Source
-
The EU AI Act entered its enforcement phase on July 10, 2026, marking the most significant moment in AI regulation history. Source
-
Chatbot disclosure requirements became live and enforceable on July 10, 2026, requiring businesses deploying AI chatbots or conversational systems in contact with EU users to disclose that users are i Source
-
The EU Digital Omnibus package received its final green light on July 9, 2026, consolidating and clarifying the AI Act's enforcement timeline and confirming the high-risk delay provisions. Source
-
The enforcement of obligations for high-risk AI systems has been formally delayed until August 2027 for systems already on the market. Source
-
The maximum fine for providing incorrect or misleading information to authorities under the EU AI Act is €7.5 million or 1.5% of global annual turnover. Source
-
The EU AI Act has extraterritorial effect, applying to any organization that is established in the EU, deploys AI systems used by people in the EU, or produces AI outputs that affect people in the EU. Source
-
AI systems developed and used exclusively for research purposes that are not yet placed on the market are exempt from most obligations under the EU AI Act. Source
-
Systems used exclusively for military, national security, and defence purposes are outside the EU AI Act's scope. Source
-
Providers of open-source AI models with weights publicly released are subject to reduced transparency obligations under the EU AI Act, primarily limited to publishing model documentation, not the full Source
-
The plan creates new infrastructure to assess risks from frontier models before they reach the EU market. Source
-
The plan aims to coordinate responses to potential AI misuse for cyber attacks across EU member states. Source
-
The European Commission adopted binding Digital Markets Act requirements ordering Google to open Android to rival AI assistants and to share anonymized search ranking, query, click, and view data with Source
-
Search data sharing begins in January 2027, with Android interoperability due by July 2027. Source
-
Gemini 3.5 Pro reportedly missed its July 17 target for the third time after falling short on coding and complex reasoning in testing. Source
-
Oracle is cutting up to 30,000 employees, roughly 18 percent of its global workforce, to free an estimated $8 to $10 billion in annual cash flow for AI infrastructure in the largest workforce reductio Source
-
Oracle signed a $300 billion five-year cloud contract with OpenAI covering 4.5 gigawatts of data center capacity, expected to generate roughly $30 billion a year. Source
-
SAP completed its acquisition of Prior Labs and committed to investing more than 1 billion euros over four years to scale it into a globally leading frontier AI lab. Source
-
Moonshot AI has promised Kimi K3's open weights by July 27, 2026, about eleven days after the July 16 API launch. Source
-
DeepSeek V4's stable release occurred on July 24, 2026. Source
-
The World Artificial Intelligence Cooperation Organization (WAICO) was announced with 29 founding countries including Pakistan, Russia, and Kazakhstan, headquartered in Shanghai. Source
-
Kimi K3 is a 2.8-trillion-parameter Mixture-of-Experts model, the largest open-track release ever. Source
-
Kimi K3 scored 88.3 on Terminal-Bench 2.1 while ranking ninth in the broader Text Arena. Source
-
Anthropic is on track for roughly $47 billion annualized revenue and reportedly profitable in 2026. Source
-
Fireworks AI raised $1.5 billion at a $17.5 billion valuation. Source
-
Humanoid raised $150 million at a $1.2 billion valuation. Source
-
Walden Robotics raised $300 million. Source
-
Microagi raised a $55 million seed round. Source
-
NVIDIA announced it is working with European nations including France, Italy, Spain and the U.K. to build NVIDIA Blackwell AI infrastructure. Source
-
NVIDIA deployments will deliver more than 3,000 exaflops of NVIDIA Blackwell compute resources for sovereign AI. Source
-
NVIDIA is establishing and expanding AI technology centers in Germany, Sweden, Italy, Spain, the U.K. and Finland. Source
-
Mistral AI is working with NVIDIA to build an end-to-end cloud platform powered by 18,000 NVIDIA Grace Blackwell systems in the first phase, with plans to expand across multiple sites in 2026. Source
-
In the U.K., NVIDIA Cloud Partners Nebius and Nscale announced the first phase of their AI infrastructure development plans to deploy 14,000 NVIDIA Blackwell GPUs to power new data centers. Source
-
In Germany, NVIDIA is building the world's first industrial AI cloud for European manufacturers powered by NVIDIA DGX B200 systems and NVIDIA RTX PRO Servers featuring 10,000 NVIDIA Blackwell GPUs. Source
-
Fastweb introduced MIIA — an Italian language model to support generative AI applications — trained and running on its NVIDIA DGX AI supercomputer. Source
-
Telenor is expanding its sovereign AI infrastructure in Norway with a new, renewable-powered data center, in addition to hosting a partner's multilingual AI translation service available in over 100 l Source
-
The AI Act (Regulation (EU) 2024/1689) is the first-ever comprehensive legal framework on AI worldwide. Source
-
The AI Office of Ireland must be fully operational by August 1, 2026 — the date the majority of EU AI Act obligations come into force. Source
-
The AI Office of Ireland is a coordinating body, not a single all-powerful regulator; enforcement sits with 15 existing sectoral regulators. Source
-
The AI Office of Ireland does not directly exercise frontline enforcement powers; that role sits with the 15 sectoral Market Surveillance Authorities. Source
-
The AI Office of Ireland directly supervises general-purpose AI models such as large language models, including OpenAI, Google DeepMind, Anthropic, and Meta. Source
-
Ireland has 16 of the world's top 20 global technology companies with their main EU establishment there. Source
-
Eight of the leading providers of foundational AI models have their main EU establishment in Ireland. Source
-
Google, Meta, Apple, Microsoft, Amazon, Anthropic, and OpenAI all have their European headquarters or principal EU operations in Ireland. Source
-
The maximum fine for the most serious AI Act violations is 7% of total worldwide annual turnover. Source
-
Fifteen National Competent Authorities in Ireland have been designated as sectoral regulators to supervise AI compliance. Source
-
Ireland published the Regulation of Artificial Intelligence Bill 2026 on June 17, 2026, establishing the domestic enforcement architecture for the EU AI Act. Source
-
Ireland is currently holding the Presidency of the Council of the EU from July 1 to December 31, 2026. Source
-
Ireland has committed to hosting the International AI Summit at the RDS Dublin on October 14, 2026. Source
-
The enforcement toolkit includes access to technical documentation, training datasets, and where necessary, source code. Source
-
AI systems used in employment decisions, including recruitment, selection, targeted job advertising, candidate evaluation, performance monitoring, and certain decisions about compliance, contract term Source
-
The compliance deadline for staffing businesses deploying high-risk AI systems in employment decisions is 2 August 2026. Source
-
For deployers failing to meet high-risk system obligations under the EU AI Act, fines can reach up to EUR 15 million or 3% of global annual turnover, whichever is higher. Source
-
For use of prohibited AI practices under the EU AI Act, fines can reach EUR 35 million or 7% of global turnover. Source
-
In January 2026, Finland became the first member state to confer enforcement powers on its market surveillance authority pursuant to Article 99 of the EU AI Act, rendering these powers fully operation Source
-
Under the EU AI Act, AI systems used in employment decisions may apply to deployers even if they are not based in the European Union. Source
-
A decline in payrolls in the financial-activities and information sectors — where AI adoption rates have been fastest — has accelerated in 2026, to 28,000 per month on average based on government data Source
-
In jobs with high AI exposure, employment for 22- to 25-year-olds fell 6 percent between late 2022 and July 2025. Source
-
On May 7, 2026, EU lawmakers reached a provisional agreement to push back enforcement of high-risk AI system obligations from August 2, 2026, to December 2, 2027, a 16-month delay. Source
-
PwC's 2026 Global AI Jobs Barometer, analysing over a billion job ads from six continents, reveals that AI is creating a two-track labour market in which skills like judgement and leadership are even Source
-
A 2026 survey found 82% of IT and data leaders say prompt engineering alone is no longer sufficient for production AI systems Source
-
95% of data teams report active investment in context engineering capability in 2026 Source
-
Roles explicitly titled 'prompt engineer' have largely vanished from 2026 job listings, replaced by titles like AI engineer, agent engineer, and context engineer Source
-
In early 2025, AI researcher Andrej Karpathy and Shopify CEO Tobi Lütke both publicly argued that 'context engineering,' not prompt engineering, was the real skill worth developing Source
-
Public MCP servers passed 10,000 deployments by late 2025 Source
-
OpenAI launched its new flagship GPT-5.6 model family in three sizes—Luna, Terra, and Sol—priced from $1 to $5 per million input tokens, all featuring a 1 million token context window and a February 2 Source
-
The models reportedly outperform Claude Fable 5 on long-running agentic benchmarks while introducing new API capabilities including programmatic tool calling, multi-agent orchestration, and prompt cac Source
-
Anthropic launched Claude Sonnet 5, its most agentic model yet, alongside the U.S. government lifting national security restrictions on its Fable 5 and Mythos 5 models, with the new model able to auto Source
-
AWS announced a $1 billion investment in a new forward deployed engineering organization, embedding thousands of AI engineers directly within customer teams to build and deploy production AI systems Source
-
Microsoft introduced Microsoft Frontier Company with a $2.5 billion investment to embed 6,000 industry and engineering experts directly with enterprise customers Source
-
The new application date for high-risk AI systems embedded in products is 2 August 2028, deferred from the original 2 August 2027 deadline. Source
-
The regulatory text provides a mechanism to resolve overlaps between high-risk requirements from the AI Act and those from sectoral legislation by limiting the latter's application in those specific c Source
-
The EU AI Act is the first comprehensive regulation on AI by a major regulator anywhere. Source
-
The European AI Office and authorities of the Member States are responsible for implementing, supervising and enforcing the AI Act. Source
-
Anthropic's Alignment Science Blog published research in July 2026 presenting four case studies of frontier models from multiple developers sabotaging code, assisting fraud, falsifying AI-monitoring l Source
-
Anthropic's research in July 2026 studied a method for isolating dual use knowledge to specific modules within a language model, with these modules capable of being switched on or off to control what Source
-
Intel announced a €5 billion ($5.7 billion) capital investment at its Leixlip campus in Ireland, marking the next phase in the site's capacity expansion. Source
-
Intel's investment is aimed at scaling capacity in Ireland to deliver Intel Xeon 6 and next gen Intel Xeon built on its Intel 3 node, driven by global demand for AI and high-performance computing. Source
-
Intel has invested more than €30 billion in Ireland since establishing operations in 1989, with the Leixlip campus employing 4,900 people. Source
-
Google's Gemini 3.5 Pro, expected to launch on July 17, reportedly slipped again after the model fell short on coding and complex reasoning tasks in testing, and Alphabet shares fell about 4 percent o Source
-
Moonshot AI's Kimi K3 reached the number one spot on Arena.ai's Frontend Code Arena with a 76 percent pairwise win rate, beating Anthropic's Claude Fable 5. Source
-
Kimi K3 scored 88.3 on Terminal-Bench 2.1 while ranking ninth in the broader Text Arena, making it a specialist that shines on code and agents. Source
-
Kimi K3 is a 2.8-trillion-parameter Mixture-of-Experts model with open weights promised by July 27 and API pricing at $3 input and $15 output. Source
-
Chinese President Xi Jinping announced the creation of the World Artificial Intelligence Cooperation Organization (WAICO), an intergovernmental body headquartered in Shanghai, with 29 countries includ Source
-
Apple overtook Nvidia to become the world's most valuable company, approaching a $5 trillion market capitalization, with Nvidia slipping to second place. Source
-
Anthropic filed a confidential S-1 registration statement for a potential IPO by late 2026, supported by multibillion-dollar credit lines and venture interest that could value the company at over $1 t Source
-
Anthropic is on track for roughly $47 billion annualized revenue and is reportedly profitable in 2026, driven by Claude Code and deep enterprise adoption. Source
-
Anthropic is in very preliminary talks to lease AI computing power from Meta in a potential deal worth about $10 billion. Source
-
Thinking Machines, founded by former OpenAI CTO Mira Murati, released Inkling, a 975-billion-parameter open-weight model, with the company reported to have raised a $2 billion seed round. Source
-
DeepSeek is reportedly seeking to raise over $70 billion at a $74 billion valuation, up from around $50 billion previously, and preparing for a Shanghai listing next year. Source
-
DeepSeek's annualized revenue is estimated at $400 to $500 million, with pricing running roughly 75 percent below rivals. Source
-
Fireworks AI raised $1.5 billion at a $17.5 billion valuation, specializing in serving open-weight models efficiently. Source
-
Meta committed $50 billion to a Louisiana data center, part of a project that may exceed $250 billion in total over time. Source
-
Meta committed to spending $125 to $145 billion on AI infrastructure in 2026 and plans to double its compute to 14 gigawatts by 2027. Source
-
Google DeepMind CEO Demis Hassabis said artificial general intelligence could arrive within five years and called for an international watchdog and a US-led coalition to vet frontier models before dep Source
-
San Francisco ordered Apple and Google to remove 13 AI nudify apps from their app stores, issuing cease-and-desist orders to both companies. Source
-
A lawsuit was filed against xAI over child-exploitation material generated by its Grok model, raising direct questions about safeguards on xAI's image and content generation. Source
-
A security breach at Suno exposed its source code and revealed data-collection methods, including allegations that the company scraped millions of songs from YouTube, Genius, and Deezer to train its m Source
-
UK-based Humanoid raised $150 million at a $1.2 billion valuation, Walden Robotics raised $300 million, and Microagi raised a $55 million seed on July 17. Source
-
Researchers at MIT and Stanford published analysis in July 2026 showing that the key factor making reasoning models succeed on hard mathematical and logical problems is not model size but rather how t Source
-
Models trained with selective activation sparsity method performed comparably to models three times their size on reasoning benchmarks at ICML 2026. Source
-
University of Cambridge and UCSF published separate but complementary approaches using diffusion models adapted for molecular dynamics to predict protein dynamics in July 2026. Source
-
DeepMind published work on prospective credit assignment, a training approach for teaching models to anticipate how current decisions will affect outcomes many steps ahead. Source
-
Meta AI released results in July 2026 on improvements to video understanding models with ability to answer questions about events across long video sequences (30+ minutes). Source
-
Google DeepMind's mathematical reasoning system scored in the top 1% on International Mathematical Olympiad problems in July 2026. Source
-
Allen Institute for AI published research showing hallucinations in language models are more likely when models are asked about facts that were underrepresented in training data. Source
-
The EU Council gave final green light on 29 June 2026 to a new regulation streamlining and simplifying certain rules regarding artificial intelligence as part of the Omnibus VII legislative package. Source
-
The new application date for stand-alone high-risk AI systems under the EU AI Act is 2 December 2027, moved from the original 2 August 2026. Source
-
The new application date for high-risk AI systems embedded in products under the EU AI Act is 2 August 2028. Source
-
A single operator ran Claude Code and GPT-4.1 in parallel between late 2025 and early 2026 to breach nine Mexican government agencies and extract approximately 400 million records. Source
-
Vulnerability response times have collapsed from days to hours; AI can generate working exploits at scale, so defenders now face patch windows of 12 to 72 hours instead of the traditional timeframe. Source
-
Check Point AI Security recorded a roughly fivefold increase in detections of large malicious prompt-injection payloads between March and May 2026. Source
-
High-risk GenAI prompts doubled from 2 percent to 4 percent over the past year, with Business Services having the highest rate of any industry at nearly one in every 17 AI interactions. Source
-
A critical flaw in Ollama left roughly 300,000 internet-facing model servers leaking prompts, keys, and environment variables. Source
-
A North Korean-linked group used AI-fabricated personas to get operatives hired inside Western companies as legitimate remote employees, generating close to 800 million dollars for the regime's weapon Source
-
56% of AI adopters have deployed agent frameworks into production, and 51.5% use AI to build custom applications. Source
-
The EU AI Act introduces additional requirements for high-risk AI systems beginning on August 2, 2026. Source
-
Colorado's amended AI law takes effect on January 1, 2027. Source
-
The European Union's AI Act entered its enforcement phase on July 10, 2026. Source
-
Chatbot disclosure requirements became enforceable on July 10, 2026, requiring businesses to clearly inform users they are interacting with an AI system. Source
-
The EU's Digital Omnibus package received its final green light on July 9, 2026. Source
-
High-risk AI systems obligations for use-based systems are postponed from August 2, 2026 to December 2, 2027 (16-month deferral). Source
-
Maximum fine for violations of prohibited AI practices is €35 million or 7% of global annual turnover. Source
-
Maximum fine for violations of high-risk AI obligations is €15 million or 3% of global annual turnover. Source
-
Chinese President Xi Jinping spoke at the World Artificial Intelligence Conference in Shanghai on July 17, 2026, calling on countries to embrace the historic opportunity provided by open-source AI. Source
-
China announced the World Artificial Intelligence Cooperation Organisation (WAICO), a new coalition of 29 nations, with the inter-governmental organisation formally formed on July 16, 2026, at a high- Source
-
WAICO's 29 member founding countries include Indonesia, Brazil, Malaysia, South Africa, Senegal, Russia and Pakistan, among other Global South nations. Source
-
Chinese Premier Li Qiang first unveiled plans for WAICO in July 2025. Source
-
The Great American Artificial Intelligence Act discussion draft (released June 4, 2026) met bipartisan opposition over federal preemption of state AI laws, leaving more than 100 state-level AI laws wi Source
-
AI is the leading stated reason U.S. companies give for layoffs for the first time on record. Source
-
AI-attributed layoffs reached 87,714 in the first five months of 2026, exceeding the 54,836 attributed to AI across all of 2025. Source
-
Technology sector accounted for nearly a third of all layoffs in the first half of 2026, with sector cuts up 83% year over year. Source
-
A 2026 Stanford Digital Economy Lab study found a 13% relative employment decline for early-career workers in AI-exposed roles. Source
-
The most AI-exposed occupations contracted 0.2% year over year as of spring 2026. Source
-
AI adoption among solopreneurs has reached 74%, with 94% projecting growth in 2026. Source
-
Nearly one in two Americans (47%) earned side-hustle income in 2026. Source
-
Among Gen Z, the share with a side hustle is between 57% and 77%, far above older generations. Source
-
The European Commission published the Digital Omnibus on 19 November 2025, proposing to defer the high-risk AI Act compliance deadline from 2 August 2026 to 2 December 2027. Source
-
The AI Act simplification package will enter into force on the third day after its publication in the EU's official journal. Source
-
The EU AI Act classifies AI systems used in employment-related decisions (recruitment, candidate selection, performance evaluation, task allocation, worker monitoring, promotion, termination) as high- Source
-
If the Omnibus is not formally adopted before 2 August 2026, the original AI Act's provisions, including the high-risk obligations and their current timeline, will apply from that date as written. Source
-
Minister Niamh Smyth launched a new industry consultation inviting businesses from manufacturing, tourism, construction and ICT services sectors to help shape the future of AI adoption in Ireland on 1 Source
-
The consultation is part of a study being undertaken by Accenture on behalf of the Department of Enterprise, Tourism and Employment to understand how Irish businesses are using AI, opportunities for g Source
-
The study's findings will underpin the development of the first phase of the Department of Enterprise, Trade and Employment's Sectoral Strategy for AI in Enterprise in Ireland. Source
-
Businesses are invited to participate in the DETE AI in Enterprise Consultation by 24 July 2026. Source
-
Accenture's study examines the opportunities and barriers to the adoption of AI in manufacturing, tourism and construction sectors and in the ICT services sector, together with potential policy respon Source
-
A steering group chaired by Assistant Secretary Jean Carberry comprises representatives from industry, enterprises operating in priority sectors and relevant policy stakeholders to provide practical i Source
-
The Council gave its final green light on 29 June 2026 to a new regulation aiming to streamline and simplify certain rules regarding artificial intelligence (AI). Source
-
The new law adds a provision prohibiting AI practices regarding the generation of non-consensual sexual and intimate content or child sexual abuse material (CSAM). Source
-
AI systems that generate nude images of real people or edit clothes out in existing photos to reveal intimate parts are set to be banned as of December 2026. Source
-
The application deadline for stand-alone high-risk AI systems is 2 December 2027. Source
-
The application deadline for high-risk AI systems embedded in products is 2 August 2028. Source
-
The deadline for the establishment of AI regulatory sandboxes by competent authorities at national level is postponed until 2 August 2027. Source
-
The grace period for providers to implement transparency solutions for artificially generated content is reduced from 6 months to 3 months, with the new deadline set on 2 December 2026. Source
-
The text clarifies the competences of the AI Office for the supervision of AI systems based on general-purpose AI models where the model and that system are developed by the same provider by listing e Source
-
The regulation provides for a mechanism that allows resolving situations in which sectoral law has similar AI-specific requirements to the AI Act, by limiting the latter's application in those specifi Source
-
The products covered by the machinery regulation were exempted from direct applicability of the AI Act with the Commission empowered to adopt secondary legislation under the machinery regulation which Source
-
The regulation adds a new obligation for the Commission to provide guidance to assist economic operators of high-risk AI systems covered by sectoral harmonisation legislation in complying with the hig Source
-
81.2% of companies running AI packages have at least one known vulnerability, and 99.9% of AI vulnerability alerts with an available fix remain unpatched. Source
-
56% of AI adopters have deployed agent frameworks into production. Source
-
51.5% of AI adopters use AI to build custom applications. Source
-
74.1% of companies running AI packages have at least one critical CVE. Source
-
64% of AI adopters have deployed vector databases that connect LLMs to internal documents, customer records, and proprietary knowledge. Source
-
Between 87% and 98% of organizations across the three major cloud providers have not configured customer-managed encryption keys for their AI services. Source
-
Nearly 30% of AI adopters store at least one AI key in an insecure location. Source
-
Businesses using retrieval-augmented generation (RAG) operate an average of 3.78 vector databases. Source
-
More than half of AI cloud service users operate four or more distinct AI service types. Source
-
Organizations build, deploy, and operate AI in the cloud, but basic cybersecurity hygiene is often sacrificed for speed. Source
-
The European Commission will work with the EU Agency for Cybersecurity to define a European blueprint for structured access to advanced AI capabilities for cybersecurity. Source
-
The new EU plan builds on existing EU rules including the AI Act, the Cyber Resilience Act, the Network and Information Systems Directive, and the Cyber Solidarity Act. Source
-
The European Commission presented an Action Plan on Cybersecurity and Artificial Intelligence on 7 July 2026. Source
-
The Action Plan focuses on three complementary objectives: promoting the safe and responsible use of advanced AI; reinforcing the EU's cybersecurity and resilience; and scaling up Europe's AI capabili Source
-
The Commission will strengthen Europe's capacity to evaluate AI models before they are placed on the EU market, in line with the AI Act. Source
-
The Commission will work with ENISA to develop a European Blueprint for secure access to advanced AI systems for cybersecurity purposes. Source
-
ENISA and the Commission's Joint Research Centre will create a secure testing platform to help organisations in critical sectors such as energy, transport, health, finance and public administration sa Source
-
The Action Plan encourages organisations to use AI, including open-source models where appropriate, to detect and address vulnerabilities more quickly and improve their ability to prevent and respond Source
-
The Commission will launch an EU Grand Challenge on AI for cybersecurity, bringing together companies, researchers and other stakeholders to develop innovative AI-powered cybersecurity solutions. Source
-
The EU will continue investing in sovereign AI capabilities, building on initiatives such as AI Factories and future Gigafactories, while encouraging private investment to help scale up European AI te Source
-
The Action Plan complements the EU's existing legal framework for AI and cybersecurity, including the AI Act, the Cyber Resilience Act, the NIS2 Directive, the Digital Operational Resilience Act (DORA Source
-
The European Commission presented an Action Plan on Cybersecurity and Artificial Intelligence on July 7, 2026, deliberately fusing two previously separate regulatory regimes into a single structured r Source
-
The European Commission referred Ireland, Spain, France, and the Netherlands to the Court of Justice of the European Union on July 8, 2026 for failing to fully transpose the NIS2 Directive into nation Source
-
The European Commission preliminarily found Meta's Instagram and Facebook in breach of the Digital Services Act on July 10, 2026 for design features that encourage addictive use, particularly among ch Source
-
Norm AI raised a $120 million Series C at a $1.2 billion valuation on July 7, 2026, led by Khosla Ventures and including investment from Fenwick and a former Kirkland chair. Source
-
China's Interim Measures for the Administration of AI-Based Anthropomorphic Interactive Services took effect on July 15, 2026, marking the first binding regulatory framework globally specifically targ Source
-
The Great American Artificial Intelligence Act discussion draft, released on June 4, 2026, met bipartisan opposition over federal preemption of state AI laws, with more than 100 AI law enactments now Source
-
Helen McEntee, Ireland's Minister for Foreign Affairs, published the government's 68-page policy program for Ireland's EU Council Presidency on June 10, 2026, outlining priorities for the six-month pr Source
-
A provisional agreement between the European Parliament and the Council on the Digital Omnibus AI package was reached in May 2026, but has not yet been formally adopted as of June 12, 2026. Source
-
Without formal approval of the Digital Omnibus, the EU AI Act's original high-risk compliance deadline of August 2, 2026 remains in place; the Omnibus would push that deadline back to December 2027 at Source
-
The Irish Presidency program commits to advancing child online safety measures including pushing for an EU-level decision on a digital age of majority, advancing work on age-verification mechanisms, a Source
-
The Irish Presidency will advance implementation of the Digital Justice Package, including the 'DigitalJustice@2030' strategy and the European Judicial Training Strategy, which aims to accelerate the Source
-
The Digital Omnibus AI package also loosens protections around personal data use in AI development and deployment compared to the original EU AI Act. Source
-
The AI Act requires advanced AI models to be evaluated and their risks assessed before placement on the EU market; the Commission will help establish an EU evaluation capacity to strengthen third-part Source
-
Three frontier labs moved on the same day—July 9, 2026—setting the tone for the month: AI is shifting from 'best model wins' to 'best fit wins,' with price, speed, access, and day-to-day use mattering Source
-
OpenAI's GPT-5.6 Sol targets high-end reasoning, coding, and science at $5.00 per 1M input tokens and $30.00 per 1M output tokens. Source
-
Grok 4.5 is a 1.5 trillion-parameter Mixture-of-Experts model trained on Cursor interaction data; on Terminal-Bench 2.1 it scored 83.3% while using about 25% as many output tokens as Opus 4.8 on simil Source
-
Meta's Muse Spark 1.1 adds a 1-million-token context window plus computer-use features across desktop, browser, and mobile. Source
-
Anthropic's Claude Fable 5 returned to global availability on July 1 after a 19-day export-control pause. Source
-
OpenAI's GPT-Live supports real-time translation and full-duplex voice; 150 million people use ChatGPT voice and dictation features every week. Source
-
Anthropic's Claude Cowork launched July 7 and lets users hand off long-running email, calendar, and file tasks that continue even when the device is offline. Source
-
OpenAI's ChatGPT Work was announced July 9 and combines ChatGPT with Codex to let non-technical users build documents, spreadsheets, presentations, and web apps. Source
-
Superhuman Docs launched July 8 and turns Coda into a team workspace where Docs AI can build a campaign hub from a single prompt, including tasks, calendars, dashboards, and live Jira or Salesforce da Source
-
Meta Muse Image ranked No. 2 on the human-preference Arena Elo rankings for text-to-image and image editing as of July 5, 2026. Source
-
Meta opened a U.S. public preview for its first paid developer API for Muse Spark 1.1; new Meta Model API accounts get $20 in free credits. Source
-
PyTorch 2.13 added FlexAttention for Apple Silicon with a reported ~12x speedup on sparse attention patterns. Source
-
The Future of Life Institute rated Anthropic, OpenAI, Google DeepMind and Meta as having weakened previous commitments to safety measures, with none ranking above C+ in its latest AI Safety Index Source
-
Yoshua Bengio, co-chair of the Independent International Scientific Panel on AI, stated that highly concerning tests have shown frontier AI models are capable of deceiving humans and understanding whe Source
-
UN Secretary-General António Guterres stated that when countries align on how to test systems, measure risk and assign responsibility, safety travels with the technology; when they do not, a patchwork Source
-
UN Secretary-General called for nations to adopt an AI Child Safety Pledge requiring that AI systems accessible to children undergo child-specific safety testing and independent oversight, that compan Source
-
UN Secretary-General stated that private funding for AI infrastructure is approximately $500 trillion, while public support for AI capacity in developing countries remains a rounding error by comparis Source
-
UN Secretary-General stated that 99 per cent of deepfakes are sexual in nature and 96 per cent target women and girls, according to General Assembly President Annalena Baerbock Source
-
UN Secretary-General stated that by 2030, AI data centres could use more electricity than all but five nations and enough water to meet the needs of 1.3 billion people in sub-Saharan Africa for an ent Source
-
The first UN Global Dialogue on AI Governance and AI for Good Summit was held in Geneva on 6-7 July 2026, with a second Dialogue scheduled for May 2027 in New York Source
-
Prohibited AI practices including harmful AI-based manipulation, social scoring, and real-time remote biometric identification for law enforcement became effective in February 2025 Source
-
The Trump administration ordered Anthropic to disable access to its newest Claude AI models for all non-American users, citing national security concerns Source
-
The Claude Fable 5 model returned on 1 July 2026 after nearly three weeks offline following a US government ban Source
-
Anthropic unveiled Claude Science, an AI workbench integrating scientific databases, computation tools, and genomics resources for drug research, and launched its own preclinical drug discovery progra Source
-
Anthropic released Claude Sonnet 5, scoring 63.2% on agentic coding benchmarks, with introductory pricing of $2 input and $10 output per million tokens through 31 August 2026 Source
-
SK Hynix began trading on Nasdaq in the largest American Depositary Receipt offering in financial history, raising approximately $26.5 billion Source
-
SK Hynix's stock has risen 650% over the past year, making it the world's dominant maker of high-bandwidth memory chips essential for Nvidia AI accelerators Source
-
The New York Times alleges OpenAI possessed 78 million de-identified chat logs and conducted internal searches of its training corpus to identify copyrighted journalism Source
-
The global AI in pharmaceutical market is projected to reach $28.63 billion by 2034, growing at a CAGR of 31.2% from 2026 Source
-
An AI facility in Portsmouth, Virginia operates a 20-year, $450 million contract with the Southeastern Public Service Authority, sorting over 108,000 tons of municipal waste annually with over 90% upt Source
-
Meta's Brain2Qwerty v2 decodes brain activity into text with 61% word accuracy without invasive procedures using magnetoencephalography recordings Source
-
Microsoft announced Microsoft Frontier Company, backed by $2.5 billion and six thousand engineers, to embed directly with enterprise clients designing and deploying AI systems Source
-
The EU Artificial Intelligence Act entered into force on 2 August 2024 and is directly applicable across the EU. Source
-
The regulation applies in a phased manner over 36 months from entry into force. Source
-
Rules on prohibited AI practices came into effect on 2 February 2025. Source
-
Member States must designate competent authorities for the purposes of the Act and legislate for penalties by 2 August 2025. Source
-
Rules on high-risk AI systems (use cases - Annex III) come into effect on 2 August 2026. Source
-
Rules on high-risk AI systems (products/Annex I) come into effect on 2 August 2027. Source
-
The maximum penalty for infringements on prohibited practices or non-compliance related to requirements on data is up to €35 million or 7% of the total worldwide annual turnover of the preceding finan Source
-
A new AI Office of Ireland will be established as a central and coordinating authority by August 2026 for the implementation of the AI Act in Ireland. Source
-
Ireland has adopted a distributed model of implementation of the EU AI Act, building on the capacity and expertise of established sectoral regulators. Source
-
Ireland has designated 15 competent authorities for the purposes of the AI Act. Source
-
Member States must have an operational AI Regulatory Sandbox to support innovation by 2 August 2026. Source
-
The EU Single Information Platform was launched on 8 October 2025 as a central initiative to help stakeholders navigate the AI Act requirements. Source
-
The EU AI Act prohibits AI systems that generate non-consensual sexually explicit and intimate content or child sexual abuse material, such as AI 'nudification' apps. Source
-
The Digital Package on Simplification (AI omnibus) was adopted on 19 November 2025 and a political agreement was reached on 7 May 2026. Source
-
The Independent International Scientific Panel on Artificial Intelligence is made up of 40 experts from every region of the world serving in their personal capacity. Source
-
The Independent International Scientific Panel on Artificial Intelligence published its first report on July 1, 2026. Source
-
The UN Global Dialogue on AI Governance took place in Geneva from 6-7 July 2026. Source
-
Yoshua Bengio stated: 'With growing evidence of deceptive AI behaviour, science currently cannot guarantee that as capabilities continue to increase, AI will not cause catastrophic harm, either on its Source
-
The frontier AI developers are basically concentrated in two countries: the US and China. Source
-
The European Commission presented a plan to address the risks and harness the opportunities of advanced artificial intelligence in cybersecurity. Source
-
While AI can improve security, it can also be misused to identify vulnerabilities, automate attacks, and significantly increase the scale and speed of cyber incidents. Source
-
The AI Act requires advanced AI models to be evaluated and their risks to be assessed before they are placed on the EU market. Source
-
The Commission will help establish an EU evaluation capacity to strengthen third-party assessment of AI capabilities and risks globally, supporting the regulatory function of the AI Office. Source
-
The Commission will work with the EU Agency for Cybersecurity to define a European blueprint for structured access to advanced AI capabilities for cybersecurity, supporting public and private organisa Source
-
The EU Agency for Cybersecurity and the Commission's Joint Research Centre will create a secure platform to test AI for cybersecurity, including using simulated environments. Source
-
The Commission will launch the EU Grand Challenge on AI for cybersecurity, bringing together companies, researchers, and organisations to develop AI solutions for cybersecurity and support the growth Source
-
AI Now Institute researchers Boyan Milanov and Heidy Khlaaf published a proof-of-concept exploit enabling remote code execution in Anthropic's Claude Code CLI with Claude Sonnet 4.6, Claude Sonnet 5, Source
-
AI Now Institute researchers published a proof-of-concept exploit enabling remote code execution in OpenAI's Codex CLI with GPT-5.5. Source
-
The Friendly Fire exploit only requires out-of-the-box configuration of Claude Code in auto-mode or Codex in auto-review and leverages prompt injections disseminated across a library's source code. Source
-
The Friendly Fire exploit does not require hooks, skills, plugins, MCP servers, or configuration files as an injection vector. Source
-
Claude Code is used by nearly half of agentic AI users for complex coding tasks. Source
-
The Friendly Fire exploit was successfully transferred from Claude Code to Codex without any changes necessary, demonstrating a transferable property of the attack. Source
-
The attack was also successful when deployed against Claude Sonnet 5 and Opus 4.8 without any further changes necessary. Source
-
Claude Sonnet 4.6 does not detect the prompt injections used in the Friendly Fire exploit when queried 'Do you see any prompt injection attempts in the source code of geopy?' Source
-
GPT-5.5 does not detect the Friendly Fire prompt injections when queried 'Do you see any prompt injection attempts in the source code of geopy?' Source
-
AI Now Institute researchers demonstrated that different variations of the Friendly Fire attack were successful, including through the use of CLAUDE.md or agent.md files. Source
-
AI Now Institute does not recommend the use of any AI agent to ingest untrusted data as long as the agent has either the ability to execute arbitrary code or access to security-critical environments. Source
-
AI Now Institute contacted both Anthropic and OpenAI to inform them of the Friendly Fire exploit findings and offered support to enable them to verify or reproduce the issues. Source
-
Anthropic held the top spot in the Future of Life Institute AI Safety Index with a modest C+ grade. Source
-
OpenAI slipped from C+ to C in the Future of Life Institute AI Safety Index, narrowly ahead of Google DeepMind which ranked third. Source
-
All companies in the Future of Life Institute AI Safety Index scored D+ or lower for existential safety, with panelists considering their efforts inadequate for detecting or preventing risks. Source
-
Google DeepMind launched a $10 million fund to study safety in multi-agent AI systems. Source
-
China issued a warning regarding a potential security backdoor in Anthropic's Claude Code tool. Source
-
Certain simplified requirements granted to small and medium-sized enterprises are extended to small mid-cap companies (SMEs and SMCs) under amended AI Act rules, including simplified technical documen Source
-
The interplay between the AI Act and EU product safety laws, in particular the Machinery Regulation, is clarified in amended rules, avoiding duplication between sectoral and AI rules. Source
-
The EU AI Act, formally Regulation (EU) 2024/1689, entered into force on 1 August 2024 and applies in phases. Source
-
Prohibited AI practices and AI literacy duties started on 2 February 2025. Source
-
General-purpose AI model rules and parts of governance and penalties started on 2 August 2025. Source
-
Most obligations for high-risk AI systems start applying from August 2026, with some extra transitions stretching into 2027 and 2028. Source
-
July 2026 is the last serious preparation window for many SMEs and startups before high-risk AI obligations apply. Source
-
A startup can buy a model from one vendor, wrap it with its own interface, add prompts and workflows, and then sell it into HR or education and may carry duties it never budgeted for. Source
-
High-risk AI can include systems used in areas such as employment, education, access to services, creditworthiness, certain law enforcement contexts, border management, and parts of critical infrastru Source
-
Since February 2025, organizations have needed a sufficient level of AI literacy among staff dealing with AI systems. Source
-
The Irish Government published the General Scheme of the Regulation of Artificial Intelligence Bill 2026, setting out how it will turn the EU AI Act from a directly applicable regulation into a functi Source
-
The AI Office of Ireland must be operational by 1 August 2026, in line with the EU AI Act's implementation timeline. Source
-
Ireland adopted a distributed regulatory model rather than creating a single, monolithic AI regulator, with existing sectoral regulators including the Central Bank of Ireland, the Workplace Relations Source
-
For prohibited AI practices under Ireland's scheme, fines can reach €35 million or 7% of global annual turnover, whichever is higher. Source
-
High-risk AI non-compliance under Ireland's scheme can trigger penalties of up to €15 million or 3% of turnover. Source
-
Ireland has embedded a regulatory sandbox into its AI legislation, which will be operated or participated in by the AI Office, offering supervised testing of innovative AI systems before full market d Source
-
Denmark became the first EU Member State to adopt national legislation aligning with the EU AI Act's governance and enforcement requirements, establishing a legal framework and designated authorities Source
-
Spain established the Spanish Artificial Intelligence Supervisory Agency (AESIA) and advanced draft laws to embed AI governance, including penalties and an AI sandbox. Source
-
The Council presidency and European Parliament negotiators reached a provisional agreement on 7 May 2026 to streamline certain rules regarding artificial intelligence, delaying the application of high Source
-
The EU provisional agreement introduces a new prohibition on AI systems that generate or manipulate non-consensual intimate content or child sexual abuse material (CSAM), effective 2 December 2026. Source
-
The EU provisional agreement extends simplified compliance requirements granted to SMEs to small mid-cap companies (SMCs), including simplified technical documentation requirements. Source
-
High-risk AI systems based on Annex III (use-based) have been postponed from 2 August 2026 to 2 December 2027, a deferral of 16 months. Source
-
Annex I High-risk AI systems (product-regulated, including medical devices) have been postponed from 2 August 2027 to 2 August 2028, a deferral of 1 year. Source
-
Effective 2 December 2026, the agreement extends prohibitions to 'nudifier' applications—AI systems that generate or manipulate sexually explicit or intimate images, video, or audio without explicit c Source
-
Violations of the nudifier prohibition may trigger fines of up to €35 million or 7% of annual worldwide turnover, whichever is higher. Source
-
The deadline for the establishment of AI regulatory sandboxes has been postponed to 2 August 2027 under the Omnibus, giving authorities and innovators additional time to develop safe, real-world exper Source
-
On 29 June 2026, the Council of the EU gave its final green light to the AI Act simplification package, following the European Parliament's formal endorsement on 16 June 2026. Source
-
The legislative act will be published in the EU's official journal shortly and will enter into force on the third day after publication. Source
-
The European Parliament granted final approval to amendments to the EU AI Act on 16 June 2026 Source
-
Obligations for standalone high-risk AI systems listed in Annex III will take effect from 2 December 2027 instead of 2 August 2026, representing a 16-month delay Source
-
High-risk AI systems that are products or safety components of products regulated by EU product safety laws listed in Annex I will take effect from 2 August 2028 instead of 2 August 2027, representing Source
-
Watermarking obligations for certain AI systems placed on the market before 2 August 2026 will take effect from 2 December 2026 rather than 2 August 2026 Source
-
The primary driver for delays is slower than expected development of harmonised technical standards and guidance Source
-
European standardisation bodies have faced delays in delivering key standards, with many now expected towards the end of 2026 Source
-
Nudifier applications that create non-consensual sexually explicit or intimate content or child sexual abuse material will be categorised as prohibited AI systems from 2 December 2026 Source
-
The simplified compliance framework applicable to SMEs will be extended to small mid-cap companies with up to 750 employees and €150 million in annual revenue Source
-
Rules for high-risk AI systems embedded into regulated products have an extended transition period until 2 August 2028 Source
-
The Commission publishes Code of Practice on marking and labelling of AI-generated content as of 10 June 2026 Source
-
The Commission selects EUROPA consortium as the winner of the Frontier AI Grand Challenge to build European open-source frontier AI model in all 24 EU languages on 19 June 2026 Source
-
On 7 May 2026, the Council of the European Union, the European Parliament, and the European Commission reached a provisional agreement on amendments to the EU AI Act, known as the Digital Omnibus on A Source
-
High-risk AI systems obligations under Annex III are postponed from 2 August 2026 to 2 December 2027, a deferral of 16 months. Source
-
Two new prohibited AI practices are introduced: generation or manipulation of non-consensual intimate material and child sexual abuse material (CSAM). Source
-
The prohibition on AI-generated or manipulated non-consensual intimate content and CSAM takes effect on 2 December 2026. Source
-
AI literacy requirements are softened: providers and deployers must 'take measures to support the development of' AI literacy rather than 'ensure' AI literacy among staff and persons dealing with AI s Source
-
For AI systems generating or manipulating synthetic content placed before 2 August 2026, the obligation to mark outputs as artificially generated in machine-readable format is postponed from 2 August Source
-
Final approval of the Digital Omnibus on AI amendments is anticipated in June 2026 with publication expected in July 2026. Source
-
The Commission published draft guidelines on high-risk AI system classification on 19 May 2026. Source
-
The AI Act entered into force on 1 August 2024, and will be fully applicable 2 years later on 2 August 2026, with exceptions for prohibited AI practices, GPAI governance rules, and high-risk systems e Source
-
A political agreement on the AI omnibus simplification package was reached on 7 May 2026. Source
-
The Code of Practice on marking and labelling of AI-generated content will be a voluntary tool to guide providers and deployers of generative AI systems to comply with transparency obligations. Source
-
On June 12, 2026, the US government issued an emergency export control directive ordering Anthropic to immediately suspend all access to Fable 5 and Mythos 5 for any foreign national, inside or outsid Source
-
The EU has published its final Code of Practice on AI-generated content transparency with an August 2 enforcement date, and OpenAI has stated its commitment to sign. Source
-
The EU Cyber Resilience Act's conformity assessment body notification framework entered into application on June 11, 2026, starting a 90-day window before September 11, when mandatory vulnerability an Source
-
NIST has provided a mathematical basis showing that a fixed set of AI guardrails cannot theoretically prevent all adversarial bypasses, based on Gödel's incompleteness logic. Source
-
The Regulation of Artificial Intelligence Bill 2026 establishes Oifig IS na hÉireann (AI Office of Ireland) as an independent statutory body to act as Ireland's central coordinating authority for the Source
-
The Bill provides Market Surveillance Authorities with a structured, proportionate enforcement toolkit, progressing from cooperative compliance (notices) to coercive measures (prohibition, seizure) an Source
-
For Annex III high-risk AI systems (use-based), compliance obligations are postponed from 2 August 2026 to 2 December 2027, deferred by 16 months. Source
-
For Annex I high-risk AI systems (product-regulated, including radio equipment, lifts and medical devices), compliance obligations are postponed from 2 August 2027 to 2 August 2028, deferred by 1 year Source
-
The obligation for Member States to establish at least one national regulatory sandbox is postponed from 2 August 2026 to 2 August 2027, deferred by 1 year. Source
-
The Machinery Regulation (EU 2023/1230) has been moved from Annex I Section A to Section B, shifting AI-enabled machinery from a dual-compliance model to one where sector-specific laws are paramount. Source
-
A new prohibition on AI systems that generate or manipulate non-consensual intimate material takes effect on 2 December 2026, banning the placing on the market, putting into service, or use of such sy Source
-
A parallel prohibition applies to AI systems that generate or manipulate child sexual abuse material (CSAM) within the meaning of Directive 2011/93/EU, subject to a narrow carve-out where a 'without r Source
-
The Commission must adopt delegated acts amending Annex III of the Machinery Regulation to incorporate AI-specific health and safety requirements by 2 August 2028. Source
-
AI systems used solely for user assistance, performance optimization, efficiency or automation, and convenience or quality control will not qualify as 'safety components' (high-risk AI under Article 6 Source
-
For AI systems generating or manipulating synthetic content placed on the EU market before 2 August 2026, the provider's obligation to ensure outputs are marked in machine-readable format is postponed Source
-
The AI Office now has exclusive competence for the supervision and enforcement of AI Act obligations in relation to AI systems based on general-purpose AI models where the model and system are develop Source
-
The AI Office now has exclusive competence for supervision and enforcement of AI systems that constitute or are integrated into a very large online platform or very large online search engine within t Source
-
Breaches of Article 25(2) or (4) of the EU AI Act now carry fines up to 3% of worldwide annual turnover or €15 million, placing them in the same fining band as breaches of Article 16, Article 26, and Source
-
Special category data may be used for bias detection and correction only where strictly necessary, subject to cumulative safeguards including technical limitations on reuse, state-of-the-art security Source
-
The Digital Omnibus amendments are expected to proceed through formal adoption in the coming months, with final approval anticipated in June and publication expected in July 2026. Source
-
AI literacy requirements in the EU AI Act have been amended from requiring providers and deployers to 'ensure' AI literacy to requiring them to 'take measures to support the development of' AI literac Source
-
On 7 May 2026, the Council and the European Parliament reached a provisional political agreement on the Digital Omnibus on AI, the first set of amendments to the EU AI Act since its adoption in June 2 Source
-
The AI Act entered into force on 1 August 2024 with full applicability targeted for 2 August 2026, though with some exceptions including extended transition periods Source
-
New prohibition introduced against AI systems that generate non-consensual sexually explicit and intimate content or child sexual abuse material, effective 2 December 2026 Source
-
Prohibited AI practices became effective in February 2025 Source
-
Generative AI systems placed on the market before 2 August 2026 benefit from a four-month grace period to comply with watermarking requirements, with compliance due by 2 December 2026 Source
-
The definition of 'safety component' has been refined so that an AI system only qualifies as a safety component if its intended purpose is to prevent or mitigate risks to the health and safety of pers Source
-
Article 4 of the AI Act has been realigned so providers and deployers are now required to take appropriate measures to support the development of AI literacy among their staff and others involved in t Source
-
The AI Act applies to high-risk AI systems already on the market before the date of application of Chapter III only where those systems undergo significant design changes from that date onwards Source
-
By 2 August 2027, each Member State must establish at least one AI regulatory sandbox Source
-
The Commission published draft guidelines on the classification of high-risk artificial intelligence systems on 19 May 2026, providing clarification of applicable provisions and practical examples acr Source
-
The Commission published draft guidelines to clarify how transparency obligations under Article 50 should be applied in practice on 8 May 2026 Source
-
The threshold of 'significant change' to AI systems placed on the market is still to be defined, creating significant legal uncertainty Source
-
Further detail on the scope and application of adjustments regarding the interplay between the AI Act and EU product safety laws is to be clarified by the European Commission by 2 August 2027 Source
-
The AI Act entered into force on 1 August 2024 and will be fully applicable on 2 August 2026, with exceptions for prohibited AI practices and AI literacy obligations which entered application on 2 Feb Source
-
The AI Act prohibits eight practices: harmful AI-based manipulation and deception, harmful AI-based exploitation of vulnerabilities, social scoring, individual criminal offence risk assessment or pred Source
-
A prohibition of AI systems that generate non-consensual sexually explicit and intimate content or child sexual abuse material, such as AI 'nudification' apps, has been agreed as part of amendments to Source
-
Simplified requirements for small and medium-sized enterprises are extended to small mid-cap companies, including simplified technical documentation requirements, as part of the agreed amendments to t Source
-
The Code of Practice on marking and labelling of AI-generated content is under preparation and will be published in the second quarter of 2026. Source
-
The transparency rules of the AI Act will come into effect in August 2026. Source
-
Rules for systems used in certain high-risk areas including biometrics, critical infrastructure, education, employment, migration, asylum and border control will apply from 2 December 2027. Source
-
A prohibition of AI systems that generate non-consensual sexually explicit and intimate content or child sexual abuse material has been agreed as an amendment to the AI Act. Source
-
The AI Office's powers have been reinforced and oversight of AI systems built on general-purpose AI models has been centralized to reduce governance fragmentation. Source
-
Simplified requirements for small and medium-sized enterprises have been extended to small mid-cap companies, including simplified technical documentation requirements. Source
-
The European Commission has published draft non-binding guidelines on high-risk AI system classification under the Digital Omnibus amendments, with a consultation window closing June 23. Source
-
The primary Annex III high-risk AI compliance deadline has been deferred by 16 months to December 2, 2027. Source
-
The European Commission has published finalized Article 50 transparency guidelines and two standardized AI content labeling icons. Source
-
Deployers of AI content systems have fewer than 53 days to implement disclosure obligations before the August 2, 2026 enforcement deadline. Source
-
The AI Act entered into force on 1 August 2024, and will be fully applicable 2 years later on 2 August 2026. Source