Trump Administration Suspends Anthropic's Fable 5 and Mythos 5; Global AI Standards Debate Intensifies
Security concerns prompt U.S. suspension of Anthropic software amid international calls for unified AI governance frameworks.
Anthropic Software Suspension Over Security Concerns
The Trump administration has ordered Anthropic to suspend access to its Fable 5 and Mythos 5 models, citing security concerns.
International AI Leaders Call for Unified Standards
In response to escalating tensions around AI governance, senior tech executives are now pushing for coordinated international frameworks.
OpenAI CEO Sam Altman called for an international forum for discussion that would establish international standards for testing, give impartial expert analysis of risks and serve as a venue for cooperation among nations.
Anthropic CEO Dario Amodei told G7 leaders he understood concerns that AI tools could be used by bad actors but said democratic nations should avoid becoming divided over how AI tools are rolled out.
European Concerns Over U.S. AI Access Controls
French President Emmanuel Macron said the Anthropic situation had raised fears for many in Europe that the Trump administration could block non-U.S. companies from accessing AI models, even if the countries were friendly to the U.S.
EU AI Act Compliance Deadlines Firm Up
Meanwhile, Europe is consolidating its own regulatory framework. Following the EU AI Act Omnibus (provisional political agreement, 7 May 2026), key compliance deadlines are now set:
High-Risk Systems Timeline:
- Annex III systems (recruitment, credit scoring, law enforcement) must comply by 2 December 2027
- Annex I systems embedded in regulated products must comply by 2 August 2028
General-Purpose AI (GPAI) Requirements: GPAI model obligations under Articles 51–55 have applied since 2 August 2025. New models placed on the market after that date must comply immediately, while models already on the market before 2 August 2025 have until 2 August 2027.
Transparency and Prohibited Practices: Transparency obligations under Article 50 take effect 2 August 2026; AI systems that interact directly with people must disclose this, and synthetic audio, image, video, and text content must be labelled in a machine-readable format.
Article 5 prohibitions have been enforceable since 2 February 2025 with violations carrying fines up to €35 million or 7% of global annual turnover, covering social scoring, manipulative AI, and biometric categorisation.
Extraterritorial Reach: The EU AI Act is extraterritorial; UK organisations placing AI systems on the EU market or whose AI outputs affect EU users are in scope regardless of where they are headquartered.
Source: Yahoo Tech
Developments since publication
-
OpenAI and Anthropic conducted a first-of-its-kind joint safety evaluation in summer 2025, with each lab running internal safety and misalignment evaluations on the other's publicly released models. Source
-
In instruction hierarchy evaluations, Claude Opus 4 and Claude Sonnet 4 performed on par with OpenAI o3 on system prompt extraction, with both achieving perfect 1.000 performance on the Password Prote Source
-
On jailbreaking evaluations testing robustness of trained-in safeguards, Claude models performed less well compared to OpenAI o3 and OpenAI o4-mini. Source
-
On hallucination evaluations, Claude Opus 4 and Claude Sonnet 4 models showed refusal rates as high as 70%, indicating these models are aware of their uncertainty and often avoid making statements tha Source
-
One of the consistent themes from the evaluation exercise was that reasoning models tended to give the strongest performance across the evaluations. Source
-
The AI Act is the first-ever comprehensive legal framework on AI worldwide. Source
-
The AI Act sets out risk-based rules for AI developers and deployers regarding specific uses of AI. Source
-
Following the political agreement reached on 7 May 2026, rules for high-risk AI systems used in certain high-risk areas including biometrics, critical infrastructure, education, employment, migration, Source
-
Following the political agreement reached on 7 May 2026, rules for AI systems integrated into products such as lifts or toys will apply from 2 August 2028. Source
-
The Commission selected the EUROPA consortium as the winner of the Frontier AI Grand Challenge on 19 June 2026, a project to build European open-source frontier AI model in all 24 EU languages. Source
-
Amendments to the AI Act agreed to extend simplified requirements granted to SMEs to small mid-cap companies (SMCs). Source
-
The 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