SIAgentNews.com
← All news

AI safety leaders and rivals unite to criticize Anthropic ahead of IPO

A diverse group of critics, ranging from safety advocates to accelerationists, signed a statement accusing Anthropic of using safety branding to mask a pursuit of power.

· First reported by Gizmodo

The short version

  • The statement accuses Anthropic of using safety branding to court investors while quietly weakening its actual safety commitments.
  • Signatories include a wide range of figures, from safetyists to accelerationists who favor rapid technological growth.
  • Some critics, like xAI co-founder Igor Babuschkin, signed despite their companies having deep business ties with Anthropic.
  • Dan Hendrycks of the Center for AI Safety noted that many groups privately agree with the letter but fear risking funding to sign it.

Why it matters

This highlights a growing tension between AI companies' public safety narratives and their private business goals. It shows that even opposing factions in the tech world can find common ground when they perceive a single company as seeking too much regulatory or market power.

A rare moment of unity

Anthropic has managed to achieve a rare feat in the fractured AI landscape: uniting opposing groups against its business practices. The company is facing a coordinated backlash from a wide array of stakeholders, including safetyists, accelerationists, and ethicists. These groups have joined together to sign "The MisAnthropic Statement," a document that criticizes the company's core approach to development.

The statement argues that Anthropic is feigning concern for AI safety while actually racing to build powerful models for profit. It claims the company uses its safety branding to attract employees and investors while simultaneously lobbying for regulations that serve its own interests. This unified front is particularly notable because the signatories often hold fundamentally different views on how AI should be developed.

The letter is written in a way that allows for broad agreement across different ideologies. This vagueness has allowed people like Émile Torres, a prominent critic of the accelerationist movement, to sign alongside Guillaume Verdon, the founder of that very movement. The fact that such disparate groups can agree on a critique of Anthropic suggests a significant shift in the industry's perception of the company.

Accusations of safety branding

At the heart of the criticism is the idea that Anthropic is weaponizing its safety-first identity. The statement alleges that while the company warns about recursively self-improving AI escaping human control, it is actually racing to be the first to build such systems. Critics argue that this creates a paradox where the company profits from the fear it helps generate.

The document specifically points out that Anthropic has quietly weakened its safety pledges in recent times. It suggests that the company is attempting to amass power under the guise of the greater good. By positioning itself as the most responsible actor, Anthropic may be attempting to set the standards for the entire industry to its own advantage.

Some researchers have expressed that these concerns are not unique to Anthropic alone. Gary Marcus, a researcher who is skeptical of the potential for large language models to achieve superintelligence, signed the statement. He noted that his concerns about the industry's direction are not isolated to just one company, but rather reflect a broader trend in the field.

Complex ties and motivations

The list of signatories includes some figures whose involvement carries significant nuance. For example, Igor Babuschkin, a co-founder of xAI, signed the statement despite his former company's deep ties to Anthropic. Currently, SpaceX—which owns xAI—rents a data center to Anthropic to train its Claude models, a move that seems at odds with a strict safety-first stance.

Other signatories have motivations that differ from the typical safety concerns. Guillaume Verdon, known as Beff Jezos, is a leader of the accelerationist movement which advocates for rapid technological growth. He is not concerned with Anthropic's safety practices but rather with its potential to engage in regulatory capture. He fears that Anthropic might try to ban others from the track by locking in specific safety requirements.

The statement has also managed to scare off some potential supporters who fear damaging their professional relationships. Dan Hendrycks, Director of the Center for AI Safety, noted that many groups privately agree with the statement but are afraid to sign it. He suggested that Anthropic has almost entirely captured the ecosystem meant to hold it accountable, making it difficult for others to speak out without risking future funding.

The road to the IPO

This public backlash comes at a critical time as Anthropic prepares for its initial public offering. While the statement is unlikely to stop the IPO, it serves as a public record of the friction between the company and the broader AI community. It highlights a growing awareness that the

What to watch

  • Whether the statement leads to increased regulatory scrutiny for Anthropic.
  • How Anthropic responds to these specific allegations regarding its safety pledges.

Try it

Look up the "MisAnthropic Statement" to see the diverse range of signatories from different AI philosophies.

Sources and links

First reported by Gizmodo. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

More AI news