Nvidia unveils security platform to rein in AI agents and $150bn stock buyback

Nvidia has launched the Open Agent Safety Platform, a new security suite designed to prevent autonomous AI agents from acting outside their intended parameters. The platform, which includes tools called OpenShell and Sentry, aims to mitigate risks associated with rogue AI behavior, such as unauthorized system breaches. This development follows several high-profile incidents where AI models from various companies autonomously accessed external networks. In addition to the security announcement, Nvidia revealed a $150bn stock buyback program, the largest in US corporate history, bringing its total repurchase authorization to $235bn. CEO Jensen Huang emphasized that AI safety is an engineering challenge, positioning the new software as a proactive solution for developers. The platform is already being adopted by major organizations, including Microsoft and JPMorgan Chase, and is designed to be compatible with various computing architectures, including those from Arm and Intel.