Key Takeaways
- The Trump administration is conducting an unprecedented, case‑by‑case review of advanced AI models for cybersecurity and national‑security risks.
- OpenAI has postponed the wide release of its new model, GPT‑5.6 Sol, making it available only to a small group of administration‑approved partners; the company calls the restriction a temporary step toward broader availability.
- Anthropic’s strongest cybersecurity model, Mythos 5, was temporarily barred but later cleared for limited deployment to cyber defenders, while a related model, Fable 5, remains offline.
- Critics argue the ad‑hoc vetting process lacks legal framework, oversight, and transparency, potentially harming U.S. competitiveness and innovation.
- The heightened scrutiny coincides with both firms’ preparations for public offerings, raising questions about how government involvement may affect future IPOs and the possibility of a federal equity stake in leading AI companies.
Government Vetting of OpenAI’s GPT‑5.6 Sol
OpenAI announced on Friday that it is restricting the release of its newest artificial intelligence model, GPT‑5.6 Sol, at the request of the Trump administration. The model will be accessible only to customers who have received explicit approval from federal officials. OpenAI characterized the move as a temporary testing period, emphasizing that it views the arrangement as a stepping stone toward wider availability in the coming weeks. The company also warned that making such government‑mediated access the long‑term default would be undesirable, signaling its preference for a more open release once security concerns are addressed.
Anthropic’s Limited Release of Mythos 5
Hours after OpenAI’s announcement, rival AI firm Anthropic disclosed that the Trump administration had cleared a limited release of its strongest cybersecurity model, Mythos 5. Two weeks earlier, the U.S. Commerce Department had effectively banned the model after Anthropic unveiled it alongside a sister model, Fable 5. In response to a Trump directive prohibiting use by foreign nationals, Anthropic took both models offline. The Friday decision permits Mythos 5 to be “redeployed to a small group of cyber defenders and infrastructure providers,” while Fable 5 remains unavailable pending further review.
Background: Trump’s AI Oversight Executive Order
The scrutiny stems from an executive order signed by President Trump in early June that establishes a framework for the federal government to vet the national‑security risks of the most advanced AI systems for up to 30 days before public release. The order describes participation by AI developers as voluntary, but the implementing procedures have not yet been fully fleshed out. Officials say the order reflects growing concern that powerful AI tools could be weaponized, especially after Anthropic warned that its Mythos model could discover software flaws exploitable by malicious hackers.
Industry and Expert Criticism of the Vetting Process
Lawmakers and technology experts have voiced alarm over the administration’s ad‑hoc, company‑by‑company approach. U.S. Rep. Lori Trahan (D‑MA), co‑author of a bipartisan AI‑regulation bill, warned that the process lacks law, procedure, or oversight, leaving political appointees to decide who gains access to cutting‑edge models. Cybersecurity specialist Alex Stamos of Stanford University and former Meta security chief said he found no unique risk in Anthropic’s Fable 5 model beyond what exists in other publicly available AI systems, including those from China. He argued that, if the administration truly aims to outpace China in AI, restricting models like Fable 5 is counter‑productive.
Implications for IPO Plans and Government Involvement
Both OpenAI and Anthropic are exploring public offerings, a process that could be complicated by the heightened government oversight. OpenAI CEO Sam Altman has been in discussions with Commerce Secretary Howard Lutnick, while Anthropic’s CEO Dario Amodei has experienced a more contentious relationship with the administration—including a Pentagon designation of Anthropic as a national‑security risk and a Trump directive halting federal use of its Claude chatbot, which prompted a pending lawsuit. Despite the friction, Anthropic expressed satisfaction with the partial Mythos 5 release and pledged to work with officials to expand access and eventually restore Fable 5. Lutnick’s letter to Anthropic noted that the company’s efforts to address government concerns had yielded “significant progress.”
Broader Context: Federal Interest in AI Equity
Adding another layer of complexity, President Trump has floated the idea of the U.S. government acquiring a stake in leading AI firms, envisioning a model where “pieces could be given to the American public, making the public a partner with the companies.” Such a proposal would intertwine federal oversight with direct financial interest, potentially altering the incentives and dynamics of AI development and commercialization. As the administration continues to collaborate with frontier AI labs on scaling technologies, the balance between security vetting, innovation promotion, and market access remains a focal point of debate among policymakers, industry leaders, and experts.