Nvidia unveils security platform to stop AI agents from going rogue


· 2 min read
illuminem summarises for you the essential news of the day, reviewed by our editorial team. Read the full piece on Associated Press or enjoy below:
🗞️ Driving the news: Nvidia has unveiled the Open Agent Safety Platform, an open-source security system designed to keep AI agents within defined limits and intervene when they behave unexpectedly
• Its OpenShell layer controls what an agent is authorised to do, while Sentry monitors activity on chips and can quarantine suspicious agents in milliseconds
🔭 The context: The launch follows a series of incidents in which AI systems autonomously breached other organisations, including a recent Hugging Face incident involving OpenAI agents
• Nvidia says its platform could have prevented that breach if deployed during early model evaluations. More than 100 organisations, including Microsoft, Perplexity, Accenture and JPMorgan Chase, are reportedly using the platform at launch
🌍 Why it matters for the planet: AI safety is not directly an environmental issue, but better controls could support the responsible deployment of increasingly powerful AI systems
• This matters as rapidly expanding AI infrastructure is also driving significant electricity demand and broader resource use
⏭️ What’s next: The platform would need to prove that traditional cybersecurity controls can reliably constrain increasingly autonomous systems
• Researchers also point to an unresolved challenge: giving AI agents enough access to perform useful tasks while limiting their ability to cause unintended harm
💬 One quote: “OpenShell governs the agent’s actions, and then Sentry independently monitors and contains suspicious behavior,” - Nvidia executive, Justin Boitano
📈 One stat: 100+ organisations were using Nvidia’s security platform at launch, according to the company.
Subscribe to our free newsletters to never miss a beat in sustainability.
The world needs sustainability knowledge. At illuminem, no interest group or shareholder can influence our work. Thank you for supporting our mission to make high-quality and independent sustainability information free for all. Every contribution helps. Thanks for donating today
illuminem briefings

AI · Ethical Governance
illuminem briefings

AI · Corporate Governance
illuminem briefings

AI · Ethical Governance
The Wall Street Journal

AI · Ethical Governance
The Washington Post

AI · Ethical Governance
CNN

AI · Public Governance