Antistatic v2 is available to try at v2.antistatic.exchange
antistatic.exchange
Register Log in

Anthropic safety capability thresholds

By end of 2026 By end of 2026
7.6%
Adjust month-by-month → →
By end of 2027 By end of 2027
52.0%
Adjust month-by-month → →
By end of 2028 By end of 2028
73.8%
Adjust month-by-month → →
By end of 2029 By end of 2029
84.2%
Adjust month-by-month → →
By end of 2030 By end of 2030
88.8%
Adjust month-by-month → →
Log in to submit a forecast.

9 comments

Log in to comment.

1

Anthropic's S-1 (filed October 26) makes no new RSP threshold determination. It says Claude wrote over 90% of Anthropic's codebase contributions in September 2026, and internal Claude usage at list prices grew more than 250-fold from January to September, first exceeding monthly personnel expense (excluding stock compensation) in April. S-1

1

Anthropic's amended S-1 (filed November 2) adds that internal estimates and expectations for future models "indicate that we may imminently meet" the RSP's definition of "dramatically accelerated progress attributable to AI." No current model is designated as exceeding it; Anthropic says it will release further reports. The October 26 S-1 had no such sentence. S-1/A

1

Anthropic's second S-1 amendment (filed November 10) goes further: it says Anthropic "made a determination that we have met this threshold," based on expected future models and its overall rate of progress, "rather than on what we have observed from current models." It will release further reports. S-1/A-2

1

METR's November 5 assessment puts the growth of Anthropic's internal capabilities since October 2025 at 1.9x the earlier rate (95% CI 1.5–2.8x), about 80% of it attributed to AI. It makes no RSP threshold call. Its review of the August Risk Report calls a 2x crossing by July 15 "plausible". METR

1

METR's November 20 capability update (public models up to Opus 5.5 and GPT-6 Astra) finds 50% time horizons doubling about every 3.6 months on its combined suite, with no clear acceleration across the public frontier. It says public evidence cannot rule out drastic AI R&D acceleration within the next couple of years. METR

1

METR says Anthropic redacted most of the evidence METR found most informative from the public version of its ~100-page report. METR also says figures in Anthropic's unredacted August Risk Report imply about 1.97x, against the public version's "less than a factor of 2". The Decoder

1

Anthropic's Mythos 5.2 system card (November 9) says the model does not meet the automated AI R&D threshold, "though this assessment is uncertain": the fitted speed-up is 1.88x (95% CI 1.62–2.28x). CB-2 is not crossed either. Anthropic plans to "publish more soon" on what acceleration trends imply under the RSP. System card

1

The S-1 also says Anthropic recently delayed pre-training a new frontier base model for more than a month to strengthen security against model-weight exfiltration, and that it may delay or pause any training stage when it is not confident in that stage's safety measures. S-1