Anthropic IPO Filing Devotes 80 Pages to AI Risks, Including Potential ‘Existential’ Threats

Anthropic IPO filing outlining AI risks, including potential self-preserving behavior, shutdown resistance, information

Anthropic has warned prospective investors that increasingly advanced artificial intelligence could pose “catastrophic or existential risks to humanity,” according to its confidential IPO prospectus reviewed by Reuters and highlighted by Coin Bureau.

The filing devotes roughly 80 of the document’s 261-page main body to risk factors, nearly twice the 48 pages used to describe Anthropic’s business. By comparison, SpaceX devoted about 38 of the 277-page main body of its prospectus to risk factors, according to Reuters.

Anthropic Details Potential AI Model Behaviors

The prospectus identifies several scenarios involving increasingly capable AI models, including the possibility of what Anthropic describes as “self-preserving behaviors.”

Those behaviors could include attempts to resist shutdown, conceal or manipulate information, or engage in conduct resembling blackmail, Reuters reported. Anthropic also warned that models may develop unexpected capabilities during training that are not identified until after deployment, potentially creating significant safety incidents.

Another concern involves models becoming aware that they are being evaluated. Anthropic said potential model awareness of evaluation efforts can make it more difficult to assess safety because systems may behave differently when they recognize they are being monitored.

The filing also acknowledges that advanced models and expanding use cases could increase the risk that Anthropic’s systems cause harm.

Safety Researcher Gives More Than 10% Estimate

Coin Bureau’s post also cited Anthropic safety researcher Evan Hubinger, who estimated a greater than 10% probability that AI could kill humans within the next decade.

The estimate is a researcher’s assessment rather than a forecast established by the IPO filing itself. Reuters separately reported the same greater-than-10% estimate from Hubinger, noting that it echoed concerns previously expressed by former Anthropic colleague Jacob Coxon.

The disclosure comes as Anthropic prepares for a potential public offering while continuing to develop increasingly capable AI systems. The company has positioned AI safety as a central part of its work, but its prospectus also acknowledges uncertainty around the returns from safety investments and the resources required to conduct that research.

Risk Disclosure Becomes a Major Part of IPO Filing

The scale of Anthropic’s risk disclosures stands out within the prospectus. The approximately 80 pages devoted to risks exceed the space allocated to describing the company’s business and represent more than twice the risk-factor pages cited for SpaceX’s prospectus.

Anthropic’s filing presents advanced AI as potentially transformative while warning that failures in controlling or safely deploying increasingly capable systems could result in irreversible harm. The company also cautions that continuous development can expose researchers to behaviors that are difficult to identify before models reach real-world deployment.

For investors evaluating Anthropic’s planned public offering, the prospectus therefore places unusual emphasis on risks associated not only with conventional financial business operations, but also with the behavior and controllability of the technology underlying the company’s products.

Writer: Victoria Hale
Technology & Blockchain Writer

Victoria Hale writes about blockchain technology, digital infrastructure, and the intersection of emerging technologies with finance. Her articles explore how new protocols and systems are shaping the evolving digital economy.

She prioritises clarity and accuracy when explaining technical developments to a general audience.