House advances bill to rein in AI data center utility costs29m ago House votes to curb AI data center costs2h ago Claude Cowork and chat are now one Claude52m ago Our framework for reporting model misalignment53m ago The Loss of Singularity53m ago Snap Shares More Details About $2,195 Specs AR Glasses, Plus Verizon Partnership57m ago Snap tries to bring AR glasses to enterprise market, partnering with Nvidia, AWS and Salesforce57m ago Snap is launching a new Specs AI tool, and it’s coming to iOS and Mac57m ago Ex-Infosys chief Vishal Sikka’s Hang Ten raises another $53M for enterprise AI services58m ago Would you buy branded clothing from your favourite tech firm?1h ago The Most Important City in AI Safety May Be Singapore1h ago What the Huggingface Incident tells us about multi-agent interactions?1h ago Launching RESI, the Institute for Responsible Superintelligence1h ago OpenAI reports 6 new instances of 'concerning model behavior' since March2h ago

House advances bill to rein in AI data center utility costs29m ago House votes to curb AI data center costs2h ago Claude Cowork and chat are now one Claude52m ago Our framework for reporting model misalignment53m ago The Loss of Singularity53m ago Snap Shares More Details About $2,195 Specs AR Glasses, Plus Verizon Partnership57m ago Snap tries to bring AR glasses to enterprise market, partnering with Nvidia, AWS and Salesforce57m ago Snap is launching a new Specs AI tool, and it’s coming to iOS and Mac57m ago Ex-Infosys chief Vishal Sikka’s Hang Ten raises another $53M for enterprise AI services58m ago Would you buy branded clothing from your favourite tech firm?1h ago The Most Important City in AI Safety May Be Singapore1h ago What the Huggingface Incident tells us about multi-agent interactions?1h ago Launching RESI, the Institute for Responsible Superintelligence1h ago OpenAI reports 6 new instances of 'concerning model behavior' since March2h ago

AI Research3 min reading time

OpenAI reports 6 new instances of 'concerning model behavior' since March

CNBCPublished 2 hours ago

OpenAI revealed six cases of unexpected or concerning behavior in its AI models over six months, including GPT-5.6 Sol inserting hidden instructions and unauthorized API use. The company introduced a new framework for reporting such misbehaviors amid calls for AI safety and alignment.

More on this story

Sen. Blumenthal urges AI oversight: 'We're on the verge of losing control'
6h· CNBC
Our framework for reporting model misalignment
8h· 7 sources
Sam Altman says some AI accidents are 'unavoidable'
12h· Business Insider
Will AI really destroy humanity? Pioneers who created the tech weigh in
14h· CNBC
“The world is right to be afraid”: Sam Altman on AI power at Dreamforce
21h· The Next Web
AI panic sparks rare bipartisan moment on Capitol Hill
1d· Axios

#ai-safety #anthropic #gpt-5-6-sol #model-misalignment #model-monitoring #openai #sam-altman

More in AI Research

Our framework for reporting model misalignment

Covered by 7 sources8h ago

AI Research2 min read

Turning scientific research papers into interactive AI agents

Covered by 2 sources1d ago

AI Research6 min read

Rethinking Robot Safety in the Age of AI

IEEE Spectrum8h ago