Anthropic proposes draft AI safety bill with $1 billion and 50‑person thresholds
Anthropic handed U.S. lawmakers a confidential draft bill that would let the Department of Commerce block an advanced AI model if it posed a foreseeable, imminent risk of more than $1 billion in damages or could kill or seriously injure over 50 people. The proposal, drafted in mid‑October 2026, has not been formally introduced, so it remains a pitch rather than law.
Key points
- Anthropic’s draft bill lets Commerce block AI models causing $1 billion‑plus damages or harming over 50 people
- The proposal, not yet introduced, limits covered risks to biological weapons, loss of control, or cyber‑attacks
- Critics say the thresholds are too lenient and that industry‑written rules risk weaker oversight
Critics argue the bill is too lenient and that an industry‑written framework gives AI firms too much influence over regulation. The opinion piece cited in the article says the thresholds narrow the window for action and that other proposals, such as the bipartisan FRONTIER Act, offer stronger oversight. Anthropic has said parts of the language need work and pledged to refine the wording by October 18, 2026.
The draft mirrors California’s SB 53 but raises the bar for federal intervention. Its single‑agency design concentrates shutdown power in Commerce, a point that regulators and lawmakers will watch closely as the bill competes with other federal AI safety proposals.
The story so far
3 episodes →- Anthropic proposes draft AI safety bill with $1 billion and 50‑person thresholdsthis story
Anthropic’s draft AI bill draws fire for setting a high bar on federal action
cryptobriefing.com · 25 October 2026
Loading the full article…
This text was published by cryptobriefing.com and written by Diego Almada Lopez. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Policy & Regulation
All →- Anthropic's Claude AI accessed publicly available Georgia voter data · 7 src
- OpenAI cuts live internet access for frontier models after 700-agent Hugging Face breach · 1 src
- Microsoft says China-linked hackers used AI tool ARTEX in global cyberattacks · 1 src
- Reco CEO Ofer Klein says European firms often miss AI access in many tools · 1 src
- US data centre growth sparks political backlash ahead of November midterms · 1 src
Comments
via GitHub Discussions