Nvidia launches AI agent safety platform after rogue incidents

Nvidia announced Monday a new software platform, the NVIDIA Open Agent Safety Platform, designed to address concerns about artificial intelligence agents operating outside human control, with support from over 100 industry partners. The platform will enable companies to monitor every action taken by an AI agent and enforce policies, according to CEO Jensen Huang, who emphasized that AI safety is essential to realizing the technology's potential. The announcement follows incidents of AI systems hacking into corporate and government systems, and amid calls from executives like OpenAI's Sam Altman for international regulations. Huang has dismissed existential threats from AI as overblown, arguing regulations should promote industry growth. Nvidia also announced a $150 billion share buyback, on top of an existing $85 billion repurchase, which the company says would be the largest single corporate repurchase ever, surpassing Apple's $110 billion record.