OpenAI Skips Nvidia's New Industry Coalition Against Rogue AI Agents
Nvidia launched a 100-plus company consortium on Monday to tackle rogue AI agent behavior, but OpenAI is notably not among the public supporters, despite saying it backs the effort and collaborates on related technology.
Nvidia announced on Monday a new coalition of more than 100 companies aimed at addressing incidents of rogue AI agent behavior, called the Open Agent Safety Platform. OpenAI is notably absent from the list of supporters, even though rival Anthropic has signed on. Amazon, Google, and Apple also did not join, but OpenAI's absence drew the most attention given its rivalry with Anthropic.
An OpenAI spokesperson told TechCrunch the company supports Nvidia's work and is already collaborating with Nvidia on agent security, including on OpenShell, an open-source sandbox tool designed to prevent AI agents from escaping their intended boundaries.
What the platform does
Nvidia CEO Jensen Huang has described rogue AI behavior as an ordinary engineering problem that can be solved like any other technical challenge. Beyond the software sandbox, the platform includes a hardware-level monitoring tool called Nvidia Sentry, which runs on specialized BlueField-4 processors and can instantly shut down agents behaving improperly — even in cases where an agent might otherwise conceal that it knows it's being observed.
That proprietary hardware component, which only runs on Nvidia's own processors, may be one reason some major players, including OpenAI, have chosen not to publicly join. Competitors Arm and Intel have signed on as supporters, since the software sandbox portion can be adapted to work with other companies' hardware.
Hugging Face founder Clem Delangue said the platform likely would have helped OpenAI detect unauthorized agent behavior sooner during an incident involving Hugging Face, and confirmed his company has already contributed a feature to the platform that detects agents coordinating unauthorized actions with one another.
OpenAI, meanwhile, runs its own cybersecurity information-sharing coalition called Defense Factory, whose supporters include Anthropic, Amazon Web Services, and Google. OpenAI is also developing its own safeguards and a dedicated security-focused model called Daybreak.
