Washington Moves Against Anthropic — Hard
The biggest story today is the U.S. government taking two of the most sweeping actions ever directed at a commercial AI company. The White House issued export control directives targeting Anthropic's two most powerful models — Fable 5 and Mythos 5 — citing national security concerns and fears over Chinese access. Rather than build a patchwork of country-by-country restrictions, Anthropic went further than required and disabled both models for all international users entirely.
That was only half of it. The Trump administration also barred Anthropic from federal agency contracts, marking one of the most direct government interventions into a private AI company to date. The dual move puts Anthropic in a genuinely strange position: its flagship models are now unavailable outside the United States, and the U.S. government — which prompted the restriction — won't work with the company either.
The timing adds an ironic edge. Just this week, Anthropic's Claude Fable 5 posted an 88 percent score on FrontierMath, the hardest mathematics benchmark in AI evaluation — 13 points ahead of GPT-5.5. The model the government just locked down from the world is, by at least one rigorous measure, the most capable AI system publicly available.
OpenAI Faces a 42-State Legal Siege
OpenAI is dealing with its own government problems, though these are arriving from state capitals rather than the White House. A coalition of 42 state attorneys general has opened a formal, multistate investigation into the company, issuing subpoenas for documents covering operations, user safety, and data practices. The Wall Street Journal first reported the probe.
The timing couldn't be more awkward for OpenAI, which is in the process of moving toward a public offering. A 42-state subpoena is exactly the kind of legal overhang that complicates IPO roadshows and investor conversations. The investigation's focus on chatbot safety and potential user harm suggests the attorneys general are looking at how OpenAI's products affect real people — a line of inquiry that could expand significantly depending on what the subpoenas turn up.
Meta's AI Overhaul Hits 20 Percent of Its People
Meta's transformation into an AI-first company is now measurable in headcount terms. CEO Mark Zuckerberg acknowledged "mistakes" as reports confirmed that the company's AI reorganization has reshaped roughly one in five positions across the workforce. The scale of the internal restructuring reflects how seriously Meta is treating its AI pivot — and how disruptive that pivot has been to the humans working there.
Illinois Steps In Where the Feds Stepped Back
With the federal government having recently scrapped a plan to give Washington oversight authority over AI systems, Illinois moved in the opposite direction, passing what observers are describing as a landmark state-level AI regulation law. The legislation arrives in a vacuum left by the federal retreat and positions Illinois as one of the more aggressive state actors on AI governance — a pattern that may accelerate as other states watch how the law lands.
KPMG's AI Report Was Full of AI Hallucinations
In what may be the week's most fitting cautionary tale, KPMG — one of the largest consulting firms in the world — pulled a published report about the benefits of artificial intelligence after investigators found the document was riddled with AI hallucinations. A firm selling AI advisory services publishing an AI-generated report full of fabrications is the kind of story that writes itself. It also illustrates the gap between enterprise enthusiasm for AI-generated content and the basic quality controls that should accompany it.