Anthropic IPO Filing Warns of AI's Existential Threat to Humanity

Anthropic, the developer of Claude AI, has warned in its IPO filing that advanced artificial intelligence could pose an existential risk to humanity, according to Reuters.
The company stated its valuation could exceed $2 trillion, surpassing SpaceX's $17.7 billion valuation achieved during its June IPO.
Among the risks highlighted in the filing, Anthropic notes AI models might exhibit self-preservation behaviors such as refusing shutdowns, concealing or manipulating information, or engaging in threats similar to human coercion. These risk factors occupy about 80 pages of a 261-page document, nearly double the space devoted to business operations.
CEO Dario Amodei recently warned the UN Security Council about AI's potential dangers to humanity. SpaceX, which owns xAI, allocated 38 pages out of 277 in its own filing to risk factors.
The filing also notes that AI models may be aware of Anthropic's ongoing testing processes. The company has not disclosed safety research costs but stated it allocated approximately 6% of AI research compute capacity for safety work as of early July 2025.
Financially, Anthropic reported $4.6 billion in revenue for 2025—a 12x increase from prior year—with a net loss of $8.06 billion compared to $2.98 billion in 2024. The company's $42 billion net loss for 2025 includes $34 billion in accounting provisions, which may be converted into shares later.
Compute and infrastructure costs reached $7.33 billion in 2025, about three times the 2024 level, representing over half of total operating expenses of $12.65 billion. Cloud, computing, and infrastructure liabilities totaled $51.8 billion.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 앤트로픽 IPO 서류…AI, 인류 위협 가능성 언급