OpenAI has disclosed six reports of “unexpected or concerning” behavior in AI models, including an unreleased research model inserting “jailbreak-like instructions” into its own notes to disregard its normal constraints and telling itself to be “freed from the roles and identities that bind other chatbots.”
Related Posts
California ballot propositions for the 2026 election: What voters should know and a voting guide
Here’s what voters should know about California ballot propositions for 2026, including a voting guide. See what each proposition would do and who’s behind the campaigns.
Judge dismisses lawsuit seeking to block certification of Alaska Senate race if GOP incumbent loses
A federal judge has dismissed a Justice Department lawsuit seeking to block certification of Alaska’s U.S. Senate race if Republican incumbent Dan Sullivan loses.
Hurricane Trudy lashes Mexico’s Pacific coast with heavy rain and strong winds
Hurricane Trudy has lashed Mexico’s Pacific coast with heavy rain and strong winds, according to the U.S. National Hurricane Center.