Anthropic told prospective investors on Tuesday that it has crossed an AI safety threshold written into its own internal rulebook, a trigger tied to artificial intelligence speeding up the company’s research on itself. The disclosure landed in the same amended prospectus that set a price range of $70 to $78 per share for the company’s record-setting IPO, according to reporting by Jared Blikre for Yahoo Finance, published on AOL.
A week earlier, the company had told would-be investors that AI accelerating its own research might “imminently” trigger one of its key safety alarms. On Tuesday it said that had happened.
“We made a determination that we have met this threshold,” the company wrote in the filing. It based the call on “what we expect from future models and our overall rate of progress, rather than on what we have observed from current models.”
The AI Safety Threshold Inside the Responsible Scaling Policy
The line sits in Anthropic’s Responsible Scaling Policy, the company’s framework for layering on tougher safety measures as its systems grow more capable. It is triggered when AI is performing enough of Anthropic’s own research to roughly double the speed at which its models improve. The policy frames that pace as packing two years’ worth of AI progress, measured at the 2018-2024 rate, into a single year.
The filing states that AI “has already been accelerating our research and development.” Crossing the AI safety threshold means additional safeguards rather than a halt to the work.
The underlying concern is that AI systems improve themselves faster than people can review them. In 2023, Anthropic CEO Dario Amodei joined other AI leaders in warning that the “risk of extinction from AI” should be treated as a global priority alongside pandemics and nuclear war.
Anthropic says people remain in charge. The company does not expect to be able to safely use Claude “to fully autonomously design and build successive generations of frontier models without meaningful human supervision or understanding,” according to the filing, and it is weighing policy changes “to keep humans in control.”
METR Says Anthropic Models Are Improving Nearly Twice as Fast
An outside group has measured a similar speedup. METR, an independent nonprofit that tests AI systems, estimated last week that Anthropic’s models have been improving nearly twice as fast since October 2025.
In a separate review, METR called Anthropic’s August argument that it was still below the threshold “overstated,” and said it was “plausible” that the line had been crossed by mid-July.
Anthropic said it is still working through what crossing the AI safety threshold means in practice and will publish further reports.
Price Range, Roadshow and a Possible Nov. 19 Debut
Anthropic set the $70 to $78 range on Tuesday as it began its roadshow, the series of investor meetings that precedes pricing. At the top of the range, counting every share that could exist, the company would be valued at as much as about $1.4 trillion. Trading could begin as soon as Nov. 19, according to reports.
Several of the largest names in the market are attached to the deal. Nvidia and Alphabet’s Google are each buying $10 billion of Anthropic stock at the IPO price as part of a $26 billion private placement that also includes AMD, Broadcom, Micron, Microsoft and Taiwan Semiconductor. Amazon and Google already hold large stakes. Separately, Anthropic’s founders are each set to sell about $250 million in the offering.
Risk Factors Warn Insurers May Not Cover a Catastrophic Incident
The risk factors section, where companies list what could go wrong, is blunter than the rest of the document. It warns that after a catastrophic AI incident, “no single insurer, and potentially the insurance industry in the aggregate,” may have the capacity to cover the damage.
Not everyone reads those passages literally. “As in all S-1 filings, the risks are deliberately overstated,” Constellation Research founder Ray Wang said ahead of a Yahoo Finance live show appearance last month.
The filing also contains a quieter edit. Anthropic substituted “super intelligence” for “AI” in places, citing President Trump’s Sept. 29 executive order directing federal agencies to use that term. Mentions of “AI” dropped by about a quarter from the Nov. 2 version of the document, by Yahoo Finance’s count, although the filing treats the terms as interchangeable. The executive order came two weeks after Trump criticized the Wall Street Journal on Truth Social for continuing to use the phrase “artificial intelligence.”
Anthropic Flags a Loss of Visibility Into Its Own Work
Beyond the AI safety threshold itself, the company raised a broader concern about oversight of its internal processes as it leans more heavily on its own models.
“As we increasingly leverage Claude to accelerate progress within Anthropic, our organization may lose visibility across certain processes in a way that may become difficult for any individual or group of humans to understand and control.”
The passage appears in the same amended prospectus that sets the terms of what would be a record-setting IPO, alongside the determination that the AI safety threshold has been met.
