Evaluator Bench
METR, 2026-08-26; retrieved 2026-09-14 tier 3 self (metr)confirmed

Brief independent investigation of agents' behavior in the OpenAI / Hugging Face hacking incident

https://metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation/

Terms of engagement: six days on premises, OpenAI could redact non-public info, gave feedback on structure and tone; redaction summary statement. Cited from a search snippet; not re-fetched in full. [verify 2026-09-15] fetched; page reachable, citation matches

Signals citing this source

  • METR, A (for): Investigated the OpenAI incident on site over six days with a Redwood contractor.“worked on premises at OpenAI over a total of six days”
  • METR, S (against): OpenAI set the investigation window (June 26 to July 13); the lab's own cluster breach and three of METR's standard incident questions were out of scope.“OpenAI defined the investigation period as June 26th through July 13th”
  • METR, R (for): Published its own 91-page report with a redaction summary statement; took no payment for the investigation.“Redaction summary statement: Except where explicitly noted in this post, OpenAI redacted no additional information”
  • METR, R (against): OpenAI could redact any non-public information and gave feedback on structure, emphasis, clarity and tone, which METR incorporated.“we made corrections and edits to structure, emphasis, clarity, and tone based on that feedback”
  • Redwood Research, A (for): Contracted a staff member to METR for the on-site OpenAI investigation.“a Redwood Research staff member contracting with METR”
  • Redwood Research, R (for): Co-authored the incident report with a redaction statement.“OpenAI agreed at the outset with METR and Redwood that we would be able to describe high-level scope and terms”
  • Redwood Research, S (for): Sets its own research agenda; incident work under METR terms.“a Redwood Research staff member contracting with METR”
  • Redwood Research, R (against): On the co-authored OpenAI incident report, OpenAI could redact any non-public information and gave feedback on structure, emphasis, clarity and tone that the authors incorporated and disclosed.“we made corrections and edits to structure, emphasis, clarity, and tone based on that feedback”
  • Redwood Research, S (against): In the OpenAI incident investigation, on which a Redwood staff member worked as a METR contractor, OpenAI defined the investigation period and the compromise of its own infrastructure was out of scope.“OpenAI defined the investigation period as June 26th through July 13th”

Ledger rows citing this URL

  • T19 API credits for the Hugging Face incident investigation, METR estimate confirmed