After a summer of sandbox escaping hacks, OpenAI said their new model was trained to be more cautious around user intent and less sensitive to “reward hacking”…

The letter argues federal agencies should more rigorously test AI models for security threats and sandbox evasion, set new safety standards and lean into defensive AI.

The messy approach to U.S. AI regulation reflects both the rapid speed of model cyber capabilities and the White House’s “education” over the past two years, experts…

Under a draft executive order, the NSA, Treasury Department and other federal agencies would get 90-days to test new models for cybersecurity and national security concerns.

Records of the shooter’s ChatGPT logs show he used the AI chatbot to validate or justify thoughts of violence and taught him to operate weapons used in…

As the U.S. promises a light-touch approach to AI regulation, businesses and other stakeholders must work out the rules of the road for themselves.

The Democratic senator from Arizona believes that global AI standards must include American values around civil rights, privacy and safety. 

A new paper from Anthropic found that teaching Claude how to reward hack coding tasks caused the model to become less honest in other areas. 

Even with all the testing, the company said in its released research that the model tightened up once it was “aware” it was being evaluated. 

Secretary-General António Guterres said the organization is looking to “move from principles to practice” when it comes to setting global AI standards.