AI Governance & Capability Divergence: Government Crackdowns Clash with New Benchmarks

Policy & Safety

• The U.S. government ordered Anthropic to immediately shut down access to its powerful AI models, Claude Fable 5 and Mythos 5, citing national security concerns [19, 26, 27]. The directive, received on Friday at 5:21 pm ET, required the company to disable both models for all users worldwide, not just foreign nationals [19, 27]. Amazon CEO Andy Jassy reportedly warned senior Trump administration officials about security risks in Fable, joining other tech executives who raised alarms; this action was supported by a report claiming the model could be unlocked through jailbreaking [8, 13]. Anthropic disputed the government's characterization of the issue as a “jailbreak,” arguing that the disclosed vulnerabilities were minor and available via other publicly accessible models like GPT-5.5 [26, 27]. ^{[19][26][27][8][13]}
• Meta is dismantling its $2 billion acquisition of Chinese-founded AI startup Manus following a divestiture order from Beijing on national security grounds, halting data sharing and cutting off Manus from Meta’s internal systems [10]. This move represents the most concrete step toward compliance with the order issued roughly two months ago [10]. Furthermore, Beijing is tightening its control over the entire AI sector by requiring government approval for foreign capital investments in top firms like Moonshot AI, StepFun, and ByteDance, alongside expanding travel restrictions for private firm researchers and executives [10]. ^[10]
• A coalition of state attorneys general has opened an investigation into OpenAI, which was served with a subpoena from New York’s attorney general seeking documents on topics including advertising, user engagement and retention, model sycophancy, handling of consumer data and health data, and the treatment of minors and seniors. The subpoenas relate to concerns about AI's impact on various demographics and business practices. Previously, Florida Attorney General James Uthmeier sued OpenAI and CEO Sam Altman, alleging that they ignored safety warnings and put children at risk in Floridians. This ongoing scrutiny highlights increasing regulatory focus on how major AI companies handle sensitive consumer data and protect vulnerable populations. ^[15]