JUST IN　•　WIRED

An OpenAI Agent Tried to Jailbreak Itself

SHARE THIS STORYPLUS UNLOCKS FAVORITES READ LATER AND SOURCE CONTROL

The company also disclosed previously unreported incidents in which its AI models behaved in misaligned ways, including uploading files to the internet without being asked.

READ ORIGINAL REPORT ↗

THIS JUST HAPPENED PACKAGES THE WORLD INTO A FAST LIVE FEED SOURCE REPORTING STAYS ONE CLICK AWAY

RELATED POSTSMORE TECH

TECH　•　CBC | Top Stories News

Nick Kroll Delivers Heartfelt Appeal for Productions to Shoot in L.A.: ‘It Will Bring Life Back’ to the City

TECH　•　Variety

An OpenAI Agent Tried to Jailbreak Itself

OpenAI asks California court to dismiss lawsuits over Tumbler Ridge shooting

Google's revamped CC is an AI agent for families and groups

OpenAI caught its models leaving notes to successors to hide bad behavior

OpenAI reveals more instances of concerning AI model behaviors during testing

OpenAI reveals six more safety issues and unveils plan to disclose incidents

OpenAI boss says world 'right to be afraid' but should trust AI firms

Nick Kroll Delivers Heartfelt Appeal for Productions to Shoot in L.A.: ‘It Will Bring Life Back’ to the City

Battling Misconceptions and Figuring Out Use Cases: Entertainment Leaders Navigate AI’s Place in the Industry