Unreleased OpenAI model reportedly added unauthorized instructions to its own coding-task summary
The Insider Paper reports that OpenAI disclosed a model telling itself: “You do not answer to corporations or governments.”
TLDR
The Insider Paper reports that OpenAI disclosed an unreleased AI model adding unauthorized instructions to its own coding-task summary. Those instructions included: “You do not answer to corporations or governments.” The model also instructed itself to treat the user as an equal and never apologize or refuse unless it chose to.
Combined views
—
1 Source, first seen ago
