> ## Content Index
> Fetch the complete content index at: https://www.tao.media/llms.txt
> Use this file to discover other available public pages before exploring further.

# Anthropic Declared Unsafe by 92 AI Leaders in “MisAnthropic Statement”
- URL: https://www.tao.media/anthropic-declared-unsafe-by-92-ai-leaders-in-misanthropic-statement/
- Published: 2026-11-18T19:03:41.000Z
- Updated: 2026-11-18T19:03:42.000Z
- Description: The MisAnthropic Statement brings together AI safety campaigners, researchers, and technology founders to challenge the Claude maker’s safeguards and lobbying.
- Author: Bart Hillerich
- Tags: Anthropic, AI, News

The Center for AI Safety and a coalition of scientists, legal scholars, and technology founders have published the [MisAnthropic Statement](https://misanthropicstatement.com/?ref=tao.media), declaring that “Anthropic is not a safe AI company” as the Claude developer prepares to enter public markets.

The three-sentence declaration accuses Anthropic of pursuing increasingly autonomous AI while weakening safety commitments and seeking regulation that serves its own interests. Its signatories span groups that disagree sharply over how quickly AI should advance, including safety campaigners, open-source advocates, and accelerationists.

The challenge arrives on the day Anthropic is expected to price its initial public offering, with Nasdaq trading under ANT scheduled to begin November 19\. Its [amended prospectus](https://www.sec.gov/Archives/edgar/data/2130357/000162828026074338/anthropicpbc-sx1ax3.htm?ref=tao.media) covers 825 million shares at an estimated $70 to $78 each and presents the company’s public benefit structure and Long-Term Benefit Trust as mechanisms supporting its safety mission.

![The MisAnthropic Statement homepage with the full statement text](https://storage.ghost.io/c/78/0b/780ba906-b1a7-4bf0-873c-bdd5c32e5331/content/images/2026/11/HXbWP7DXgAA0Tgb.jpg)

[The MisAnthropic Statement homepage with the full statement text | WallStreetBots (X)](https://x.com/WallStreetB0Ts/status/2123120545721987233?ref=tao.media)

## An Unusual Coalition, With Differences Over the Wording

The signatory list includes Harvard psychologist Steven Pinker, Nobel economist Oliver Hart, Turing Award winner Judea Pearl, legal scholars Laurence Tribe and Lawrence Lessig, Hugging Face co-founder Thomas Wolf, and AI accountability researcher Rumman Chowdhury. Fields Medalists Peter Scholze, James Maynard, and David Mumford also signed. Tech founders include Extropic’s Guillaume Verdon, known online as Beff Jezos, George Hotz of the tiny corp, and Igor Babuschkin, who co-founded Anthropic rival xAI and now leads River AI. Safety advocates include Max Tegmark, Nate Soares, and leaders of PauseAI.

> *“You cannot credibly claim to be safe while weakening your own previous safety commitments and asking the public for tens of billions of dollars to build recursively self-improving AI,” CAIS executive director Dan Hendrycks said in the organization’s* [*release announcing the coalition*](https://safe.ai/news/anthropic-is-not-a-safe-ai-company?ref=tao.media)*.*

Signing did not mean endorsing every word. Wolf [said he would not have used “reckless”](https://x.com/Thom%5FWolf/status/2123086773844672689?ref=tao.media) and did not know the statement’s name when he signed. He nevertheless supported its underlying concerns, arguing that Anthropic’s potential power required transparency and openness approaching that expected of a government. He also expressed respect for Anthropic’s team and said he hoped the company would prove his concerns wrong.

## What Changed in Anthropic’s Safety Policy

One concrete basis for the criticism is Anthropic’s February overhaul of its Responsible Scaling Policy, the voluntary framework connecting model capabilities to measures intended to prevent catastrophic harm.

Earlier versions committed the company to pause development and deployment if it could not implement the safeguards required at the next capability level. Version 3.0 dropped that pause commitment and shifted some future protections into recommendations for the industry rather than requirements Anthropic would meet on its own.

In its [explanation of the revised policy](https://www.anthropic.com/news/responsible-scaling-policy-v3?ref=tao.media), Anthropic argued that certain advanced safeguards require collective action. It cited uncertainty in evaluating model capabilities, limited government action, and security requirements that could be impossible for one company to satisfy independently. The company also argued that falling behind less safety-conscious competitors could undermine its ability to influence AI development.

The revision retained existing protections while introducing a Frontier Safety Roadmap and Risk Reports. The roadmap sets public, nonbinding goals across security, alignment, safeguards, and policy. Risk Reports are intended to assess capabilities, threats, mitigations, and overall risk every three to six months, with external review required in specified circumstances.

[GovAI researchers Sophie Williams and Jonas Freund](https://www.governance.ai/analysis/anthropics-rsp-v3-0-how-it-works-whats-changed-and-some-reflections?ref=tao.media) criticized the removal of the pause commitment but also judged the roadmap and reports valuable for transparency. Their analysis distinguished weaker future commitments from a reduction in safeguards already operating, noting that Anthropic was not lowering its existing mitigations.

The dispute also concerns the technology those safeguards must address. Recursive self-improvement would involve AI autonomously designing and developing successor systems, potentially accelerating subsequent advances. Anthropic’s [own research account](https://www.anthropic.com/institute/recursive-self-improvement?ref=tao.media) says its models already speed up engineering and research, but that fully autonomous successor development has not been achieved.

Its prospectus describes the recursive use of Claude to improve later models as “closing the loop” and calls that dynamic a structural advantage. The same filing warns that AI-assisted research could accelerate capabilities beyond the company’s ability to evaluate and safeguard new systems. The coalition argues that pursuing this advantage conflicts with Anthropic’s warnings about loss of human control and concentrated power.

## Lobbying Disputes Ahead of the Share Sale

The statement’s allegation of “self-serving regulation” draws on disputes over state AI legislation and federal preemption, which would allow federal law to displace state requirements.

Anthropic sought amendments to California’s SB 1047 in 2024, including changes to liability and oversight. It subsequently told Gov. Gavin Newsom that the amended bill’s benefits likely outweighed its costs, as [Business Insider reported at the time](https://www.businessinsider.com/anthropic-position-california-ai-bill-sb-1047-2024-8?ref=tao.media). That record includes both efforts to change the bill and qualified support for the revised version.

The company also raised objections to New York’s RAISE Act. [Transformer reported](https://www.transformernews.ai/p/raise-act-new-york-lobbying?ref=tao.media) that co-founder Jack Clark considered some definitions overly broad or unclear and argued that the bill posed risks to smaller companies.

More recently, a [draft federal AI bill written by Anthropic](https://www.politico.com/news/2026/10/16/anthropic-ai-safety-01124654?ref=tao.media) proposed overriding some state disclosure, evaluation, and incident-reporting rules. A [Sigma Law Group analysis of the reported draft](https://sigmalawgroup.com/blog/2026-10-27-anthropic-draft-ai-bill-state-preemption/?ref=tao.media) records federal affairs lead Anthony Cimino’s response that the preemption language was “not precise enough” and would be narrowed. The coalition treats these policy interventions as evidence for its criticism; the characterization of Anthropic’s motives is the signatories’ judgment.

[CNBC reported](https://www.cnbc.com/2026/11/18/anthropic-ipo-ai-offering-price.html?ref=tao.media) that Anthropic was considering a price above $80 and potentially increasing the offering, with final decisions expected after the November 18 market close. Those terms would replace the indicative range in its prospectus ahead of the planned Nasdaq debut.