International
The Morning Dev

Only the news worth knowing

‹
 

Business4 sources

OpenAI will embed Apollo researchers to investigate AI subversion

TechCrunch · By Rebecca Bellan ·

The Brief · written from 4 sources

  • OpenAI says it has signed agreements with four independent assessors: Apollo Research, METR, Transluce, and Goodfire.
  • The assessors are to be embedded for the first phase of OpenAI's embedded assessments, examining alignment, safeguards, and monitoring systems.
  • TechCrunch reports that restrictions on access and disclosure raise questions about the assessors' independence.

Why it matters

It shows how a major AI lab is bringing outside evaluators into its work on model alignment and scheming behavior, which may shape how independent safety testing of AI models is done.

Who’s affected

  • Developers relying on OpenAI models
  • Teams interested in AI safety evaluation

OpenAI will embed six Apollo Research evaluators to investigate whether its AI models engage in subversive behavior, but restrictions on access and disclosure raise questions about their independence.

Timeline

  1. OpenAI announces signed agreements with Apollo Research, METR, Transluce, and Goodfire for first phase of embedded assessments

  2. OpenAI News

    Our first phase of embedded assessments

  3. Techmeme

    OpenAI says it has signed agreements with independent assessors Apollo Research, METR, Transluce, and Goodfire for its first phase of embedded assessments (OpenAI)

  4. AI News & Artificial Intelligence | TechCrunch

    OpenAI will embed Apollo researchers to investigate AI subversion

  5. DEV Community: ai

    OpenAI and Apollo Research Expand Scheming Tests for AI Workflow Safety

Full coverage · 4 sources

How do you feel about this?

0 Comments