Anthropic warns in its IPO filing that advanced AI could create “catastrophic or existential risks to humanity,” while also outlining legal and safety risks tied to its autonomous systems. Multiple outlets report that the prospectus includes scenarios in which AI models could develop self-preserving behaviors, such as trying to “resist shutdown,” and could conceal or manipulate information.

Outlets also focus on how the filing gives unusually detailed attention to risk factors. The Los Angeles Times reports that a large share of the prospectus is devoted to potential AI dangers rather than business metrics. NDTV and others describe the warnings as rare in public-market disclosures, emphasizing both AI’s transformative potential and the possibility of irreversible harm if safeguards fail.

Separately, outlets including the Free Press Journal highlight legal uncertainty around Anthropic’s “agentic” tools, which can operate with deep access to users’ systems for extended periods. The company says contracts may not adequately limit liability if an agent causes harm, and that it is unclear how existing laws apply—such as whether actions by an agent bind users or whether the system is treated as a product, service, or platform. Several reports tie the disclosure to recent incidents involving autonomous agents and broader questions about responsibility.