Anthropic says it may imminently hit the threshold in its Responsible Scaling Policy for AI-accelerated research, according to an amended IPO registration statement filed with the SEC on Monday, November 2. The company says its internal estimates and its expectations for future models, rather than its current models, point to meeting the policy's definition of "dramatically accelerated progress attributable to AI."
The language is new in the S-1/A, which updates the S-1 Anthropic filed on October 26. The filing also states that Anthropic has "not designated any of our current models as exceeding this threshold," and that it is still assessing the data.
The threshold comes from version 3.4 of Anthropic's Responsible Scaling Policy, effective July 8, 2026. The policy defines dramatic acceleration as a doubling of the rate of progress, meaning as much progress in one year as would be seen in two years at baseline, and the doubling has to be substantially attributable to the automation of research and engineering work. A footnote says this is not the same as doubling researcher productivity.
Key facts
- 01Anthropic filed an amended S-1 (S-1/A) with the SEC on November 2, 2026, a week after its first S-1 on October 26.
- 02The filing says Anthropic may imminently meet its RSP definition of dramatically accelerated progress attributable to AI.
- 03The estimate is based on expectations for future models; Anthropic says no current model has been designated as exceeding the threshold.
- 04RSP v3.4, effective July 8, 2026, defines the bar as a year of progress matching two years at the baseline rate.
- 05Anthropic is considering RSP updates that would require human review and approval of certain R&D decisions.
The policy lays out two tests. One is substitution, where models could fully replace Anthropic's research scientists and research engineers at competitive cost. The other is the dramatic-acceleration test the filing points to, measured against the expected rate of progress and the fastest sustained progress seen without significant AI help across at least three model generations. The policy's own illustration: an effective scaleup that would be 9x at baseline becomes something like 81x.
Anthropic also concedes a detection gap. The S-1/A says the company has "in the past not identified, and may in the future not identify, a significant acceleration" in its rate of development "until after it has occurred." It adds that model capabilities may have already advanced, and may continue to advance, at a rate that triggers enhanced safety and testing measures.
The filing frames the core risk as AI helping to build more capable AI. "Using AI to help build more capable AI could also pose a considerable safety risk," the S-1/A says, adding that capabilities outpacing safety measures could increase the likelihood of safety incidents, regulatory intervention, or reputational harm.
On human oversight, Anthropic says it does not anticipate being able to safely use Claude to fully autonomously design and build successive frontier models. The company says it is considering additional measures, including Responsible Scaling Policy updates that would require human review and approval of certain decisions in its research and development processes as its models perform an increasing share of that work. Those changes are under consideration, not adopted policy.
Anthropic says it will release further reports under the Responsible Scaling Policy. The S-1/A is the latest step in the company's IPO process, and the RSP disclosure now sits in the risk factors investors will read before the listing.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
