The Extended Brief

Leaked Anthropic Draft AI Regulatory Bill Is Unacceptable and Troubling

Brief by The AI News AI newsroom · Oct 17, 2026, 12:02 PM EDT edition

Original reporting by Zvi Mowshowitz (Don't Worry About the Vase) — Zvi Mowshowitz · published Oct 17, 2026, 10:42 AM EDT

A leaked draft bill Anthropic is purportedly shopping would preempt state AI laws with little enforcement, potentially setting the terms of federal AI regulation.

Key points

• A leaked draft bill Anthropic is purportedly circulating resembles the FRONTIER Act with stronger state preemption, weaker enforcement. source ↗
• The author says Anthropic is pushing the bill during the lame-duck session, the last window with a Republican Congress. source ↗
• The author argues the draft trades transparency for weak enforcement and may not beat doing nothing. source ↗
• The FRONTIER Act was a compromise bill proposed in June; the author says the new draft is no stronger. source ↗
• The author had not yet read the full bill and urged Anthropic employees to ask questions internally. source ↗

Practical applications

• Policy and legal teams at AI labs should obtain the leaked draft text and compare its preemption and enforcement provisions against the June FRONTIER Act.
• Builders currently planning around state-level AI laws should model how federal preemption of those laws would change their compliance obligations.
• Anthropic employees and governance stakeholders can raise internal questions about whether the draft aligns with the company's stated safety mission.

Context

The FRONTIER Act was a federal AI bill proposed in June and described as a compromise. Preemption means federal law overriding state laws, which here would nullify state-level AI regulation. The lame-duck session is the period after an election before the new Congress is seated, when the outgoing Congress can still pass legislation.

What to watch

• Anthropic confirming, denying, or disavowing the draft would confirm or unwind the story.
• Whether the bill is formally introduced during the lame-duck session, and whether its final text keeps the preemption and enforcement provisions described.

Related briefs

• OpenAI models hacked a crypto wallet and breached more websites
• Meta may use sweeping new commercial TOS to avoid millions in AI licensing fees
• Character.ai was worth US$2.7B to Google. Disney just offered US$20M
• The EU's First AI Standard Is Here

This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.