Anthropic says S1 may soon cross RSP threshold for AI-accelerated research progress
Anthropic's recent S1 filing told investors the company may soon satisfy its Responsible Scaling Policy trigger for dramatically accelerated progress driven by AI systems themselves. Under the RSP revision effective July 2026, that bar is tied to observing or expecting overall capability advances at roughly twice the fastest sustained pace from before AI contributed materially to AI R&D. Separately, an Alignment Forum and LessWrong analysis by Ajeya points to METR colleagues' deep-access work suggesting Anthropic's internal capabilities index already shows about a twofold speedup between October 2025 and October 2026 compared with the prior year. The author argues the dramatic-acceleration threshold may already be crossed, though that reading rests on a proprietary metric and is not an official Anthropic declaration. Builders and policymakers should treat the gap between investor-language proximity and a confirmed policy tripwire as the central uncertainty.
Anthropic says S1 may soon cross RSP threshold for AI-accelerated research progress
Anthropic told investors in its S1 that it may imminently meet its Responsible Scaling Policy bar for dramatically accelerated progress driven by AI. Under the July 2026 RSP, that tripwire is tied to observing or expecting overall capability progress at about twice the prior pace.
Key takeaway
A reported RSP threshold crossing from METR-style measurement is not the same as Anthropic formally declaring the tripwire met in its S1 language.
What happened
Anthropic told investors in its S1 that it may imminently meet its Responsible Scaling Policy bar for dramatically accelerated progress driven by AI, according to reporting summarized on the Alignment Forum and LessWrong.
An Ajeya post on those forums cites METR deep-access work finding capabilities progress at Anthropic from October 2025 to October 2026 was about twice as fast as the prior twelve months, measured via an internal Epoch Capabilities Index variant called AECI, and concludes the RSP dramatic-acceleration threshold has already been crossed.
Evidence
Anthropic's S1 said it may imminently meet the RSP threshold for dramatically accelerated progress due to AI.
Alignment Forum · attributed
Earlier this week, Anthropic stated in their S1 that they "may imminently meet" the threshold of "dramatically accelerated progress due to AI" as defined in their Responsible Scaling Policy.
The July 2026 RSP ties the threshold to capability progress going about twice as fast.
Alignment Forum · attributed
In their latest RSP (effective July 2026) , say this threshold will be met when: They observe or expect overall capabilities progress to be going twice as fast
METR deep-access work reported roughly doubled capabilities progress at Anthropic between October 2025 and October 2026.
LessWrong · attributed
Between October 2025 and October 2026, capabilities progress at Anthropic has been going about twice as fast as it had been over the previous twelve months.
The acceleration measurement used an internal Epoch Capabilities Index variant called AECI.
LessWrong · attributed
The measurement they report is an internal version of the Epoch Capabilities Index called AECI
Why it matters
If AI-driven acceleration of AI R&D is treated as a live RSP threshold event rather than a future scenario, urgency sharpens around scaling commitments, evaluation transparency, and how builders and policymakers respond.
Limits and uncertainties
The threshold-crossing conclusion is a secondhand interpretation of an internal METR report and RSP wording, not a confirmed Anthropic declaration.
The twice-as-fast claim rests on proprietary AECI construction; methodology and error bars are not fully detailed in the excerpt.
Practical implications
Teams tracking RSP-style commitments should separate S1 proximity language from formal tripwire declarations and independent index readings.
Builders should prioritize scrutiny of capabilities measurement definitions before treating dramatic acceleration as an established milestone.
What to watch
Whether Anthropic formally declares the dramatic-acceleration RSP threshold met beyond S1 may-imminently-meet language.
Public detail on METR deep-access methodology and robustness of the AECI twofold speedup finding.
