US issues executive order targeting Anthropic's models, opening a new era of AI oversight

The US government has issued an official order directly concerning specific models from Anthropic, the developer of Claude. Bloomberg describes the document as the start of a new era in artificial intelligence regulation: for the first time, US federal authorities are targeting an individual developer rather than the technology industry as a whole.

What happened

The document, reported by Bloomberg, introduces a new regime of government oversight over Anthropic's AI models. At the center of the report is a project codenamed Mythos, whose nature is still disputed. It is either a classified version of Claude built for federal agencies, or a separate mechanism for jointly managing powerful AI systems under government control.

Bloomberg discloses the order's specific terms only in part. The overall conclusion, however, is unambiguous: a precedent has been set. The US government now has tools to directly regulate the commercial AI systems of a specific developer, and this is not a mere recommendation but a binding official act.

Why Anthropic

Anthropic was not chosen at random as the subject of the first such order. The company holds a distinctive place in the AI industry:

• It develops some of the most powerful frontier models, the Claude line and its successors
• It consistently positions itself as a "responsible AI lab," which raises regulators' expectations
• It has received strategic investment from Google and Amazon, making it systemically important to the American tech economy
• It actively takes part in government AI safety initiatives, including the National AI Safety Institute and related programs

This combination of technological weight, corporate ties and public stance on safety may have made Anthropic a convenient "pilot case" for testing a new regulatory mechanism.

The Mythos puzzle

Mythos is the key element of the report. Bloomberg explicitly calls the story "convoluted" and does not reveal all the details. Depending on interpretation, it could be:

• A closed version of Claude developed at the request of US military or intelligence agencies
• An internal Anthropic program for monitoring and restricting potentially dangerous capabilities of its models
• A joint government-corporate mechanism for real-time oversight of frontier AI systems

Whatever is hidden behind the codename, the very fact that it appears in a federal document matters: AI regulation in the US is moving beyond abstract declarations.

"This order signals the arrival of a new era in AI control," —

Bloomberg Tech.

What it means

The US has long avoided strict federal AI regulation, preferring voluntary agreements with developers and advisory standards. This order changes that logic: the government is starting to work directly with specific systems and specific companies, not with the industry as a whole.

For Anthropic, OpenAI, Google DeepMind and other frontier AI leaders, the signal is clear: the era of setting safety rules for themselves is coming to an end, and formal government oversight is replacing it.