METR says Anthropic redacted most of the evidence it found most informative in its ~100-page report
METR says it submitted a roughly 100-page full report on its assessment to Anthropic, and that Anthropic redacted a majority of the evidence METR found most informative from the public version. METR says readers must therefore rely on its judgment more than it considers ideal.
METR says Anthropic redacted a majority of the evidence it found most informative from the public version of a roughly 100-page report. The statement was posted by METR on X (@metr_evals) and surfaced on Techmeme on November 5, 2026.
What METR says
According to METR, it submitted a ~100-page full report on its assessment to Anthropic. For the public version, METR says, Anthropic redacted "a majority of the evidence we found most informative."
METR states the consequence directly: "the public must rely on our judgment more than we consider ideal." In the headline summary of the post, METR describes the redacted material as the "most informative" evidence, and says readers "must rely on our judgment more than we consider ideal."
Key details
- Author of the statement: METR, via its @metr_evals account on X.
- Length of the full report: approximately 100 pages, as stated by METR.
- Party that redacted: Anthropic, according to METR.
- Scope of redaction: a majority of the evidence METR considers most informative, according to METR.
- Stated effect: readers of the public version depend more heavily on METR's own conclusions than METR considers ideal.
These details come from METR's account. Only a short feed summary of the post was available, so further specifics of the assessment are not covered here.
What this means
A public report is only as checkable as the evidence it contains. METR says the redactions removed most of the evidence it considers most informative, so outside readers have less underlying material to test its conclusions against. METR itself frames this as a departure from the ideal: its judgment, rather than the full evidence, carries more of the weight in the public version. Readers evaluating the report should treat its conclusions as METR's assessment, and should note that METR says the fuller evidence behind them exists in a ~100-page version that was submitted to Anthropic.
Related Articles
METR Report: Anthropic's AI Development Pace May Have Doubled Due to Internal AI Use
A new report from AI research nonprofit METR says AI use within Anthropic may have doubled the company's pace of AI development. The rate at which Anthropic's models are improving as of October is roughly twice the rate between March 2024 and October 2025, according to The Information.
Anthropic Warns Small AI Models Are Months From Self-Replicating Botnets; Experts Call Threat Familiar
Anthropic's Frontier Red Team published research suggesting small AI models may be only months away from replicating across insecure devices as autonomous botnets. Experts told Engadget the capability is plausible, but that the framing overstates how new the threat is.
AI Running Costs Fell This Year Even as Server Prices Rose, but Data Center Curbs Test the Trend
The cost of running AI has fallen this year even as server rental prices rose. After U.S. voters elected officials who want to restrict data center expansion, The Information reports that some executives and investors worry capacity delays could stall that decline.
Hirundo and Lazarus AI edit weights of Chinese open models to strip censorship for Western clients
Startups Hirundo and Lazarus AI are editing the weights of Chinese open-weight models, including Alibaba's Qwen, to remove state-aligned censorship and harden security for Western enterprises, Rest of World reports. Some researchers question the premise, and some users see no need for it.
Comments
Loading...