Unconfirmed: an AI Alignment Forum post1 says METR found Anthropic’s AI R&D progress had accelerated by October 2026, but Automatica has not confirmed the underlying assessment.
The post, titled “METR: Independent Assessment of AI R&D Acceleration at Anthropic in October 2026,” was published on 5 November on the AI Alignment Forum1. Its executive summary says, “From August 12 to October 29, 2026, METR staff performed a high-access evaluation at Anthropic” and that Anthropic offered “unprecedented access to internal data on their AI R&D process,” including pretraining and reinforcement-learning metrics, training compute by domain, benchmark scores for internal models, AI speedup analyses, forecasts and staff interviews AI Alignment Forum1.
According to the same summary, the authors “now believe” that, as of October 2026, Anthropic was “likely improving the capabilities of their models on well-specified tasks at around twice the rate they did between March 2024 and October 2025,” while adding “high uncertainty” and saying “the true acceleration was closer to 1.3x or 3x the historical rate” would not be surprising AI Alignment Forum1. The post also says “a majority of this acceleration today is likely attributable to Anthropic’s extensive use of AI to do R&D,” though it adds “significant uncertainty” and says other causes likely mattered in late 2025 AI Alignment Forum1.
The summary further says that “the lack of observed bottlenecks or strong diminishing returns means there is a significant risk that further capabilities progress could lead to even faster acceleration,” while cautioning that the authors “may be missing important bottlenecks” AI Alignment Forum1. It also gives a more specific metric: “Since October 2025, benchmark scores (measured via AECI, Anthropic’s version of the Epoch Capabilities Index) for Anthropic’s models have been growing at an estimated 1.9x the pre-October 2025 rate [95% confidence interval: 1.5–2.8x]” AI Alignment Forum1.
The post refers to a “Full Report” that Automatica has not seen. That means the methods, baseline choices, data and full argument behind those figures are not available for independent review here, and neither METR nor Anthropic appears to have published a separate statement in the material reviewed.
What would confirm the rumour is publication of that full report, or a direct statement from METR or Anthropic confirming the assessment, the scope of access and the reported acceleration figures. What would cut against it would be a correction, withdrawal or contradiction from METR or Anthropic.
Automatica reported earlier on 5 November that an X post claimed METR had obtained unusual access to Anthropic’s internal model work and seen results from unreleased systems, while noting there were then “no primary documents or independent corroboration” [Automatica Press earlier coverage2]. The forum post is more specific than that earlier claim because it gives dates, a described evaluation window and numerical estimates, but it remains a public post whose central claims have not been independently confirmed.