80 pages of risk. The 261‑page prospectus, which dwarfs the 48‑page business description it previously used, signals that Anthropic is prioritizing safety over immediate revenue streams.
In a stark admission, safety researcher Evan Hubinger flagged a probability exceeding 10% that AI could kill humans in the next decade—a figure that sits alongside the company’s own cautionary stance.
Meanwhile, the company disclosed that roughly six percent of the compute power it deployed for research last July was earmarked for safety work, underscoring the resource‑intensive nature of its protective measures.
Compared to SpaceX, which allocated 38 pages to risk factors in its 277‑page prospectus, Anthropic’s 80‑page commitment suggests a more aggressive risk narrative.
The filing also indicates that Anthropic plans to release more data on model usage and safety protocols in the coming months, hoping to assuage investor concerns while maintaining a rapid release cadence.
Investors have responded with caution, trading the company’s shares in the pre‑market session at modest gains, reflecting the market’s wary stance toward the high‑risk, high‑reward AI landscape.