MisAnthropic Statement
Confirmed
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 3 of 5
- Last verified
- 19 November 2026
Your AI and this story
- GPT-6.2 Astra171 days after its cutoff
- Claude Opus 5.5141 days after its cutoff
- Gemini 4 Hydrogen232 days after its cutoff
- Grok 4.7171 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 141 days before it.
Key facts
- Statement: ‘Anthropic is not a safe AI company’; it says the company ‘uses safety branding to court investors, employees, and the public’ while it ‘races to build’ recursively self-improving AI, and asks the public and policymakers to ‘reject’ its bid to amass power (CAIS press release, Nov 18)
- Evidence CAIS cites (its characterisation): the February 2026 Responsible Scaling Policy revision that dropped unilateral pause commitments; lobbying against SB 1047, the RAISE Act and the AEGIS Act and for federal pre-emption of state AI laws; pitching recursive self-improvement to IPO investors
- Signatories named by CAIS: Dan Hendrycks (CAIS), Steven Pinker, Oliver Hart (Nobel, economics), Judea Pearl (Turing Award), Lawrence Lessig and Laurence Tribe (Harvard Law), Thomas Wolf (Hugging Face), Igor Babuschkin, Rumman Chowdhury, Tristan Harris, Guillaume Verdon (‘Beff Jezos’), Max Tegmark, George Hotz, Geoff Ralston, Fields medallists Peter Scholze, James Maynard and David Mumford; full list at misanthropicstatement.com
- The coalition joins safety advocates with accelerationists and open-source figures (Verdon, Hotz, Wolf) who usually oppose them; HN thread 134 points
- Ban RSI rally (banrsi.ai), outside Anthropic’s offices on Nov 19 at 11 am; Daniel Kokotajlo (AI Futures Project) said he would attend: ‘Recursive Self-Improvement is a stupidly dangerous idea… IMO there are no good guys in this race’ (X, Nov 18, ~98k views). Protesters also gathered at the Nasdaq MarketSite on listing day and some were arrested for blocking the doors (CNBC)
- Same days (headlines): FT, ‘Anthropic’s AI advances trigger new safety fears on eve of IPO’; NYT, Dustin Moskovitz says he ‘invested in Anthropic to promote safety, not racing’; The Verge, Anthropic’s ‘good guy’ reputation is on the line; WSJ live blog, ‘AI-safety advocates challenge Anthropic’s safety branding’; The Information, Craig Newmark holds back AI-safety donations because of the IPO
- Anthropic’s own position (S-1/A, Nov 10): it has determined it met its RSP threshold for AI-accelerated R&D and says it intends to pace frontier development (see Anthropic: AI R&D threshold met (S-1/A))
What happened
CAIS released the statement with a press release and a signature site the day Anthropic priced its IPO. It argues that a company cannot claim to be safe while weakening earlier commitments and raising tens of billions of dollars to build AI that improves itself. The claims about Anthropic’s lobbying and policy changes are CAIS’s characterisations.
Why it matters
It is the broadest public challenge yet to Anthropic’s safety identity, and it brought together people who usually disagree about AI risk. It frames recursive self-improvement, which several labs now openly measure and discuss, as the line the signatories want drawn.
Sources
6 sources from 6 sites. Numbers match the chips in the text.
6 sources: 1 primary, 3 press, 2 reactions
Primary
- CAIS: Anthropic is not a safe AI company (press release, Nov 18)safe.ai, official
Press
- Business Wire: ‘Anthropic is Not a Safe AI Company’ – broad coalition challenges Anthropic’s safety brandingbusinesswire.com, press
- The Verge: Anthropic’s ‘good guy’ reputation is on the linetheverge.com, press
- FT: Anthropic’s AI advances trigger new safety fears on eve of IPO (headline)ft.com, press
Reactions
- Daniel Kokotajlo on X: Ban RSI rallyx.com, discussion
- Hacker News discussionnews.ycombinator.com, discussion
Changes
- Filed