Anthropic Says AI Safety Alarm Has Been Triggered in IPO Filing
Anthropic says it has crossed a safety threshold tied to AI accelerating its research, citing expectations for future models and the company’s overall pace of progress. The disclosure came in an amended IPO prospectus that also set a share price range of $70 to $78.
Anthropic said on Tuesday that its AI safety alarm had been triggered, a disclosure in an amended IPO filing that also set a share price range of $70 to $78. The company said it based the decision on expected future models and its overall pace of progress, rather than what current models have demonstrated.
The threshold concerns AI systems doing enough of Anthropic’s research to roughly double the pace at which its models improve. Crossing it calls for stronger safeguards, not a shutdown.
“We made a determination that we have met this threshold,” Anthropic wrote in the filing.
The alarm is part of the company’s Responsible Scaling Policy, which sets out additional safety measures as its AI systems become more powerful. Anthropic said AI “has already been accelerating our research and development.”
The policy describes a pace that would compress two years of AI progress, measured at the rate from 2018 to 2024, into one year. Anthropic’s judgment relied on what it expects future models to do, not a finding that current models have already reached that level.
The company has said it does not expect to use Claude to independently design and build successive generations of advanced models without meaningful human oversight. It is considering policy changes “to keep humans in control.”
METR, an independent nonprofit that tests AI systems, estimated last week that Anthropic’s models had been improving nearly twice as fast since October 2025. In a separate review, METR called Anthropic’s August argument that it remained below the threshold “overstated” and said the line may have been crossed by mid-July.
Anthropic began investor meetings on Tuesday and set its IPO range at $70 to $78 a share. That range implies a valuation of as much as about $1.4 trillion, counting all shares that could exist. Trading could begin as soon as Nov. 19.
Nvidia and Google are each set to buy $10 billion of Anthropic stock at the IPO price, as part of a $26 billion private placement involving AMD, Broadcom, Micron, Microsoft and Taiwan Semiconductor. Amazon and Google already hold large stakes.
The prospectus also warns that a catastrophic AI incident could exceed insurance capacity. It says “no single insurer, and potentially the insurance industry in the aggregate,” may be able to cover the damage.
Anthropic said it is still assessing what crossing the threshold means and will publish further reports. The filing also warned that greater reliance on Claude to accelerate the company’s progress could reduce its visibility into some processes.
