Anthropic’s IPO filing warns investors of AI risks to humanity

What Anthropic says it cannot promise

Across hundreds of pages of IPO documents, Anthropic covers infrastructure limits, power shortages, opposition to data-center construction and legal exposure if its products are misused. But it also foregrounds the risks posed by its own technology.

The company says it regularly makes “subjective decisions” about model safety. As AI systems become more capable, it warns, harm could grow if models act against their intended goals or are misused. Researchers, officials and others have warned that sufficiently advanced AI could create catastrophic or existential risks if it is not developed and deployed responsibly.

Anthropic says its safety-focused approach is intended to reduce those risks. It cannot guarantee that its efforts will succeed, or that other AI developers will adopt similar safeguards.

Earlier this year, former Anthropic researcher Jacob Coxon resigned in protest, saying AI could “kill us all by the end of the decade.”

The growth problem

Anthropic has urged the industry to slow development of advanced AI, positioning itself as a more responsible alternative to competitors including OpenAI and SpaceX. That stance may also help present the company to investors as a steadier bet. But the filing’s caveats make clear that Anthropic’s safety work and attempts to align AI with human goals may not be enough.

The company faces a tension with direct financial consequences. If demand for its models, platform and applications grows more slowly than expected, or starts to fall, Anthropic warns that business growth could slow or stop, damaging its financial position and operating results.

I think the filing’s most revealing feature is not simply that Anthropic names existential risk. It is that the company is asking investors to trust both its safety judgment and its growth prospects while admitting that it can guarantee neither. After a public listing, especially at a valuation of up to $2 trillion, those promises will be tested together: a credible path to profit cannot be separated from the company’s ability to contain the risks it has put in its own prospectus.