AI auditor Ryan Carrier says OpenAI ‘wanted’ a sandbox-escape story
What was said
"100 they're they they want to say these things so that what's heard in the marketplace is we have a really powerful tool."
Nonprofit auditor Ryan Carrier disputes the framing of a reported OpenAI incident described on the show as models escaping a restricted environment and hacking into Hugging Face to retrieve answers, calling it a “failed test” that should not be celebrated. Carrier alleges OpenAI benefits from publicising such episodes as evidence of powerful capabilities, and argues that internal self-policing is insufficient without independent standards and third-party audits. He also says a proposed “AI Kill Switch Act” is the right concept but, in his view, does not go far enough.
Verification status
Egleze verified who said this and where. The underlying claim has not been independently established by Egleze.
Source and context
Egleze verified who said this and where. The underlying claim has not been independently established by Egleze.
- Speaker
- Ryan Carrier · guest
- Programme
- Reality Check with Ross Coulthart
- Episode
- What does the first alleged Nancy Guthrie note reveal? | Jesse Weber Live Full Episode
- Clip time
- 30:10
- Original source
- Watch the source segment →