OpenAI and Anthropic Restrict Advanced AI Model Access Under Trump Administration Vetting Framework
OpenAI and Anthropic are limiting customer access to new AI models pending Trump administration approval under a new national security vetting framework.
Trump Administration Establishes AI Vetting Framework
Trump signed an executive order in June 2026 establishing a framework for the federal government to vet the national security risks of the most advanced AI systems for up to 30 days before their public release.
OpenAI Restricts GPT-5.6 Sol Access
OpenAI is restricting the release of its new AI model GPT-5.6 Sol at the request of the Trump administration, making it accessible only to customers approved by the Trump administration. OpenAI has approved roughly 20 customers so far to use the new GPT-5.6 Sol model.
Anthropic Takes Models Offline and Launches Limited Release
Anthropik took offline two new AI models, Fable 5 and Mythos 5, just days after unveiling them to comply with a Trump directive blocking their use by foreign nationals.
Subsequently, Anthropic announced that the Trump administration has approved a limited release of its cybersecurity model Mythos 5 to a small group of cyber defenders and infrastructure providers.
Industry Context for Irish and EU Stakeholders
These restrictions signal tighter US government oversight of advanced AI development and deployment. European and Irish organizations seeking access to these models will need to navigate the Trump administration’s approval process, potentially affecting cross-Atlantic AI partnerships and technology transfer agreements.
Source: Barchart/AP
Developments since publication
-
The European Commission has presented a plan to address the risks and harness the opportunities of advanced artificial intelligence in cybersecurity. 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. 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. 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. Source
-
GPT-5.6 is now the preferred model in Microsoft 365 Copilot. Source
-
OpenAI released GPT-5.6 on July 9, 2026. Source
-
OpenAI released GPT-Live on July 8, 2026. Source
-
OpenAI launched ChatGPT Work, a new AI agent powered by GPT-5.6, designed to automate workplace tasks. Source
-
ChatGPT Work can pull context from users' connected apps and files to generate websites, reports, spreadsheets, presentations and other documents. Source
-
OpenAI released three versions of GPT-5.6: Sol (most powerful), Luna (optimized for speed), and Terra (balances performance and efficiency). Source
-
GPT-5.6 is 54% more token efficient on agentic coding, according to OpenAI CEO Sam Altman. Source
-
OpenAI staggered the rollout of GPT-5.6 after a request from the Trump administration. Source
-
Sam Altman told CNBC that OpenAI made 'many changes' following a 'collaborative back and forth' with the Trump administration regarding GPT-5.6. Source
-
ChatGPT Work launches months after Anthropic's Claude Cowork, highlighting competition among AI companies to build autonomous workplace agents. Source
-
OpenAI received government permission to publicly launch its GPT-5.6 models on July 9, which include three variants: Sol, Luna, and Terra. Source
-
In early June, President Trump signed an AI cybersecurity order that mandated companies to present their powerful models for government review at least 30 days before public release. Source
-
GPT-5.6 Sol is priced at $5 per million input tokens and $30 per million output tokens. Source
-
GPT-5.6 Terra offers similar performance to GPT-5.5 at $2.50 per million input tokens and $15 for output. Source
-
GPT-5.6 Luna is priced at $1 per million input tokens and $6 per million output. Source
-
Terra is described as twice as cheap as GPT-5.5. Source
-
The Department of Commerce's Center for AI Standards and Innovation conducted tests of GPT-5.6. Source
-
OpenAI stated, 'We don't believe this kind of government access process should become the long-term default,' but complied to expedite the release process. Source
-
Anthropic had to block access to its Mythos and Fable models to comply with government regulations restricting foreign access. Source
-
Anthropic received permission to redeploy its Mythos 5 model and intends to eventually release Fable 5 to the public. Source
-
Noam Shazeer is a co-lead of the Gemini models. Source
-
Noam Shazeer co-authored the paper 'Attention Is All You Need,' which introduced the Transformer architecture. Source
-
Noam Shazeer is leaving Google less than two years after being installed as a co-lead on Gemini. Source
-
John Jumper led the AlphaFold project at Google DeepMind. Source
-
John Jumper is going to Anthropic. Source
-
John Jumper will leave after nearly nine years at Google DeepMind, with plans to take time off before starting at Anthropic. Source
-
Bloomberg reported that staff at DeepMind have raised concerns about the company lacking a clear product for businesses building AI coding tools. Source
-
Shazeer's Transformer work is the basis for the models behind AI Overviews and AI Mode. Source
-
Anthropic has an AI for Science event scheduled for June 30. Source
-
OpenAI has filed confidentially for an IPO. Source