Trump drops restrictions on Anthropic's Mythos and Fable models | TechCrunch

What Happened

According to a TechCrunch report, the Trump administration has lifted restrictions that had been placed on Anthropic's Mythos and Fable models, removing a regulatory obstacle that had constrained the release or deployment of these systems. The report frames this move as part of a broader pattern of unpredictable AI policymaking coming out of Washington, one that has left companies across the industry guessing about what rules — if any — will apply to future model releases.

Why This Matters

The decision underscores a defining feature of the current US approach to AI governance: volatility. Unlike the European Union, which has codified its expectations through the EU AI Act and is now moving into the harder work of enforcement — audits, risk classifications, and penalties for noncompliance — the United States continues to govern AI largely through executive discretion. Restrictions can appear and disappear depending on the administration's priorities, leaving AI labs like Anthropic without a stable compliance roadmap.

For Anthropic specifically, the removal of restrictions on Mythos and Fable likely accelerates its ability to ship or expand these products without the regulatory friction that had been holding them back. But the broader signal is arguably more consequential than the specific models involved. If restrictions can be imposed and then reversed based on shifting political winds rather than durable statutory frameworks, companies are incentivized to lobby the executive branch directly rather than build for compliance with a fixed set of rules. That dynamic tends to favor well-resourced incumbents who can navigate — or influence — Washington, while smaller players face more uncertainty.

The Bigger Regulatory Picture

This episode also feeds into ongoing debates about antitrust and market concentration in AI. When safety and deployment rules are applied unevenly or renegotiated behind closed doors, it raises questions about whether regulatory decisions are being shaped by competitive considerations as much as by safety analysis. Critics of light-touch, ad hoc AI oversight argue that this approach risks entrenching the market power of a handful of frontier labs — including Anthropic, OpenAI, and Google DeepMind — by making regulatory outcomes negotiable rather than predictable.

Looking Ahead

The contrast with the EU's more codified, if slower-moving, enforcement regime is likely to sharpen in the coming months. As the AI Act's provisions come into force, American companies operating internationally will need to comply with Brussels' rules regardless of what Washington decides domestically. Meanwhile, without clearer congressional action, US AI policy will likely keep oscillating with each administration, a pattern that leaves both safety advocates and industry executives without the certainty they say they need.