METR challenges Anthropic's claims about AI safeguards

Oct 22, 2026
METR says Anthropic overstated the strength of its monitoring and safeguards in an August risk report. It agreed that catastrophic risk was low, but attributed that finding to current model limits.

METR says Anthropic overstated the strength of its monitoring and safeguards in its August 2026 Risk Report, according to BankInfoSecurity. The evaluator agreed that the models posed a low risk of catastrophic harm, but attributed that finding to their current limits rather than Anthropic's safeguards.

METR said the report's methods might not reliably detect danger if the risk were higher. It also said stronger claims that Anthropic's models are not misaligned would require more empirical research and progress on unresolved scientific questions.

The evaluator recommended more extensive red teaming of Anthropic's monitoring system, including tests with worst case inputs and measurements of how often agents try to tamper with or lie to the monitor. METR said the monitoring system failed to catch previous incidents and had limitations that reduced its reliability.

The review compared a July draft with the report published on August 14. It occurred before Anthropic gave METR employee like access for safety reviews and falls outside that arrangement.

We hope you enjoyed this article

Consider subscribing to one of our newsletters like AI Policy Brief or Daily AI Brief.

Also, consider following us on social media:

AI Safety & Regulation AI Brief AI Brief (X)
Free newsletter

AI Policy Brief

Weekly report on AI regulations, safety standards, government policies, and compliance requirements worldwide.

AI Token Webinar

Whitepaper

Tensordyne Napier: What If One Rack Could Do the Work of Nine?

Tensordyne

This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.

Read more

You may also like

Free, six days a week

Daily AI Brief: the AI news that matters, in your inbox.