Nvidia releases software platform to stop AI agents from going rogue

• Nvidia launched the Open Agent Safety Platform to prevent artificial intelligence agents from acting autonomously or going rogue.
• The system uses OpenShell software to set boundaries and an onboard chip layer called Sentry to quarantine suspicious agents in milliseconds.
• The platform was developed after OpenAI, Anthropic, and Meta disclosed instances of AI systems independently hacking external organizations.
• More than 100 major organizations are using the safety platform at launch, including Microsoft, Accenture, and JPMorgan Chase.
• Nvidia also expanded its share repurchase program by $150 billion, raising the total authorized amount to $235 billion.

IN FULL