MisAnthropic Statement: Pearl, Tegmark, Marcus and Others Say Anthropic Is ‘Not a Safe AI Company’ Ahead of IPO

A coalition including the Center for AI Safety, Judea Pearl, Max Tegmark, Gary Marcus and Thomas Wolf published the MisAnthropic Statement, arguing that Anthropic uses safety branding while racing to build recursively self-improving AI, days before its IPO.

Christopher Ort
By Christopher Ort · 3 min read
Share
Printed letters with signature lines and a pen on a conference table.
Image: AI-generated, i10X
On this page
  1. Quick Take
  2. In plain English
  3. What happened
  4. Background
  5. What the coalition argues
  6. Why it matters
  7. What's still open
  8. Sources
  9. i10x Perspective

A coalition of AI safety researchers, open-source advocates, scientists and legal scholars published the MisAnthropic Statement on 18 November 2026, days before Anthropic’s expected stock market debut. The short statement argues that Anthropic uses safety branding while racing to build recursively self-improving AI, and it is backed by the Center for AI Safety and signatories including Judea Pearl, Max Tegmark, Gary Marcus and Hugging Face co-founder Thomas Wolf.

i10X Quick Take

Quick Take

  1. What: a public statement at misanthropicstatement.com, announced by the Center for AI Safety (CAIS) in a press release.
  2. Core claim: “Anthropic is not a safe AI company, and we must reject its reckless attempt to amass power in the name of the greater good.”
  3. Signatories include: Steven Pinker, Oliver Hart, Judea Pearl, Laurence Tribe, Lawrence Lessig, Thomas Wolf, Igor Babuschkin, Rumman Chowdhury, Tristan Harris, Guillaume Verdon, Max Tegmark, George Hotz, Gary Marcus, Nate Soares and Fields medalists Peter Scholze, James Maynard and David Mumford.
  4. Cited evidence: CAIS points to the February changes to Anthropic’s Responsible Scaling Policy and the company’s lobbying positions on state AI bills.
  5. Timing: Anthropic’s IPO is pending; its last filed price range was $70 to $78 per share.

In plain English

Anthropic has built its reputation on being the careful, safety-first AI lab. A broad group of well-known researchers and public figures now says that image does not match what the company does. They argue Anthropic warns about AI that can improve itself, yet is racing to build exactly that and promoting it to investors. The statement is meant to influence how investors and the public see Anthropic as it goes public.

What happened

The statement opens: “Anthropic warns that recursively self-improving AI could concentrate power and escape human control, yet races to build it first.” It says the company “uses safety branding to court investors, employees, and the public” even as it “has quietly weakened its safety pledges and lobbied for self-serving regulation,” and ends: “Anthropic is not a safe AI company, and we must reject its reckless attempt to amass power in the name of the greater good.”

CAIS executive director Dan Hendrycks said in the release: “People in this coalition disagree on almost everything, yet we all recognize the danger of a company that convinces itself its own dominance is the greater good.”

Anthropic had not publicly responded to the statement at the time of writing.

Background

Anthropic describes itself as an AI safety and research company. Its IPO filings have included extensive risk disclosures about advanced AI and large compute commitments. In February 2026 the company revised its Responsible Scaling Policy, the framework that ties model releases to safety safeguards. Debates over recursive self-improvement, where AI systems speed up their own development, have grown this year, including calls for limits on the practice.

What the coalition argues

  • Racing toward recursive self-improvement: Anthropic warns about it in public while presenting it to investors as a competitive advantage, according to CAIS.
  • Safety branding: the label is used to recruit safety-minded staff and to pitch investors as the company works to automate AI research.
  • Weaker commitments: CAIS says the February policy change dropped Anthropic’s unilateral commitments to pause, and that the company lobbied to weaken SB 1047, the RAISE Act and the AEGIS Act and privately backed federal pre-emption of state AI laws.

Why it matters

For investors, safety reputation is part of Anthropic’s pitch to enterprises and governments. A coordinated challenge from prominent scientists lands at the moment that pitch is being priced.

For the AI safety field, the coalition is unusual: safety researchers, e/acc figures such as Guillaume Verdon and open-source advocates rarely sign the same document.

For Anthropic’s customers, nothing changes in the products today, but the statement raises questions about how much weight to give lab-written safety frameworks compared with independent oversight.

What’s still open

  • Whether and how Anthropic responds before its listing.
  • Whether more researchers or organisations sign; the website accepts new signatures.
  • How investors weigh the criticism against reported strong demand for the offering.

Sources

  • Center for AI Safety, “Anthropic is Not a Safe AI Company,” press release, 18 November 2026
  • The MisAnthropic Statement, misanthropicstatement.com, 18 November 2026

i10x Perspective

Every frontier lab now faces the same question: can a company that sells ever more capable models also be trusted to set its own limits? Anthropic tied its brand more tightly to that answer than its rivals did, so it draws the sharpest criticism. For background, see our coverage of the risks disclosed in Anthropic’s IPO filing, the Anthropic company hub and the Policy & Safety section.

Christopher Ort

Christopher Ort

Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU). He is the author of “The Style Penalty”, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.

Timeline

What happened before

  1. Release

    Claude Haiku 5.5

    1M context · $0.1 / $0.5 per 1M tokens

    support.claude.com (source, opens in a new tab)
  2. Release

    Claude Sonnet 5.5

    1M context · $2 / $10 per 1M tokens

    support.claude.com (source, opens in a new tab)
  3. Release

    Claude Opus 5.5

    1M context · $4 / $20 per 1M tokens

    First model of the Claude 5.5 family; performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5.

    anthropic.com (source, opens in a new tab)
  4. Partnership

    Accenture partnership on independent frontier-AI evaluation (≥$1B each over five years)

    With Accenture · Research

    anthropic.com (source, opens in a new tab)
  5. Release

    Claude Fable 5.1 & Mythos 5.1

    1M context · $10 / $50 per 1M tokens

    Anthropic's most advanced models for coding and knowledge work. Price/context shown for Claude Fable 5.1 (models overview).

    support.claude.com (source, opens in a new tab)
Full Anthropic timeline ›

Editor's picks

  1. GAO Report: ChatGPT, Claude, Gemini and Grok Produced Viable Attack Plans After Safeguards Were Bypassed
  2. ResolveAI SRE-1: A 27B Model Trained to Find the Root Cause of Production Incidents
  3. Reson8 Rumble-2-Flash: Open-Weight 0.5B Speech Recognition Model Covers 40+ Languages
  4. Pika Enact Lets Creators Direct AI Video Actors and Cameras in Real Time
  5. Google Stitch Loop: An Always-On Product Agent That Fixes Security, Speed and Accessibility via Pull Requests