Reaction
OpenAI says it is bringing outside safety evaluators into internal model testing
OpenAI said it is working with Apollo Research, METR, Transluce, and Goodfire to assess internal models.
TLDR
Apollo Research will test whether models could covertly work against developers’ intentions and whether safeguards would detect that behavior, OpenAI says. The company says assessors receive scoped access, while it retains the right to redact confidential information and request publication delays for safety or security risks.
Combined views
—
2 Sources, first seen ago
Comments