Nvidia Thinks It Can Stop Rogue AI—Without All That Government Oversight

President Donald Trump listens as Nvidia CEO Jensen Huang speaks about US investments at the White House on April 30, 2025.Alex Brandon/AP Get your news from a source that’s not owned and controlled by oligarchs. Sign up for the free Mother Jones Daily. On Monday, Nvidia, the multinational tech company—which has become one of the most valuable companies in the world in significant part thanks to its role in developing chips for AI processing and model training—announced the launch of a new safety platform that will monitor and contain AI agents, amid recent reports of “rogue AI” incidents like OpenAI’s Hugging Face hacking fiasco. As my colleagues and I have previously reported, accounts of autonomous hacks by AI models belonging to OpenAI and its rivals have exacerbated concerns about the minimally regulated AI sector, even as leading figures like Nvidia CEO Jensen Huang argue that AI oversight is an “engineering problem” that the industry can largely handle without government involvement—echoing statements by President Trump, who has lashed out at demands to legislate guardrails around what he now calls “superintelligence.” Nvidia says its Open Agent Safety Platform, which it launched Monday with a number of leading tech and AI firms, can quarantine agents that attempt to escape containment within “milliseconds.” Among company’s collaborators on some of its new safety measures are Microsoft, Palantir, SpaceXAI, JPMorganChase, and Hugging Face—the online forum where users can share and build AI systems that was the subject of the hacking incident involving OpenAI. Nvidia says its Open Agent Safety Platform, which it launched Monday with a number of leading tech and AI firms, can quarantine agents that attempt to escape containment within “milliseconds.” The company’s collaborators on some of its new safety measures include, among others, Microsoft, Palantir, SpaceXAI, JPMorganChase, and Hugging Face—the online forum where users can share and build AI systems that was the subject of the hacking incident involving OpenAI. But in the context of Huang’s remarks, the company’s proposals for developing safe AI are in effect being presented as an alternative to regulatory action. The idea that AI safety isn’t a problem for state agencies, or that the recent spate of rogue AI incidents don’t call for swift political action, amount to an argument that companies like OpenAI and Anthropic can police themselves. It’s often AI firms’ own conduct, and the words and actions of their leaders, that have made self-regulation a hard sell—especially as public opinion sours towards the technology’s sprawling infrastructure demands and footprint in daily life. Disclosure: The Center for Investigative Reporting, the parent company of Mother Jones, has sued OpenAI for copyright violations. OpenAI denies the allegations.