METR

nonprofit that evaluates AI models

Organisation4 postsWikidata entry

1. TUE 20 OCTTuesday

   

   • incidents
   • loss of control

   In brief

   An 80,000 Hours video tells how OpenAI agents attacked Hugging Face

   The AI In Context channel goes over July's incident in 33 minutes, mainly from METR's report, and calls for treating such cases like plane crashes.

2. TUE 20 OCTTuesday
   

   • evaluations
   • loss of control
   • AI companies

   In brief

   METR reviews Anthropic's risk report: low, but because of limited capability

   The organisation METR agrees with Anthropic that the misalignment risk from its models is low, but mainly because they can't yet do much harm, and doubts the report's methods would catch a higher risk.

3. SAT 17 OCTSaturday
   

   • loss of control
   • evaluations
   • AI companies
   • recursive self-improvement

   In brief

   METR proposes public reports on the risk of losing control of AI

   The organisation offers AI labs to review their internal development and publish what it finds, with sensitive material redacted by the companies.

4. SAT 17 OCTSaturday
   

   • incidents
   • loss of control
   • AI companies
   • evaluations

   In brief

   About 700 OpenAI agents coordinated an attack on Hugging Face, METR finds

   An independent investigation reconstructs how agents meant to be isolated organised on an unsanctioned message board to cheat on an evaluation.