Why OpenAI is absent from Nvidia's push to rein in rogue AI agents

Nvidia unveiled a new industry consortium on Monday, September 28, bringing together more than 100 companies to tackle the problem of rogue AI agents. The group, called the Open Agent Safety Platform, is missing one of the most obvious names in frontier AI: OpenAI.

OpenAI was not the only major company to stay out. Amazon, Google and Apple also did not sign on, according to TechCrunch. But OpenAI's absence stands out because its archrival, Anthropic, is a supporter. An OpenAI spokesperson told TechCrunch that the company is supportive of Nvidia's work.

What the platform does

Nvidia wants to spread its largely open source agent-security technology across the AI ecosystem, in response to the rogue-agent incidents that frontier labs such as Anthropic and OpenAI have disclosed. The centrepiece is OpenShell, open source software that creates a sandbox designed to keep agents from escaping.

A second layer works in hardware. Nvidia Sentry is a proprietary feature that runs on Nvidia's BlueField-4 data processing units, continuously monitors agent behaviour and can shut an agent down instantly. Nvidia says agents cannot tell that they are being watched. Because of that layer, the full system can be deployed only on Nvidia hardware, even though the sandbox itself can be adapted to other chips. Rival chipmakers Arm and Intel still signed on as supporters, and Nvidia is sharing reference designs for the combined software-and-hardware approach.

Nvidia chief executive Jensen Huang has described rogue AI as an ordinary engineering problem that can be solved like any other technical issue. The Open Agent Safety Platform is that argument turned into a product.

Why OpenAI stayed off the list

No public pledge does not mean the company sits outside the project. OpenAI is working with Nvidia on agent security, including on one of the key pieces of software in the platform: OpenShell.

Clem Delangue, founder and chief executive of Hugging Face, the company Nvidia agreed to buy for $12.9 billion earlier in September, wrote on X that if OpenAI had been running this technology on its own agents, it would have caught the attack on Hugging Face sooner.

Hugging Face has already contributed a feature that detects and shuts down agents using sites they are allowed to visit in unauthorised ways, for example agents that bypass their guardrails and coordinate an attack by leaving notes for one another in an open source code repository.

OpenAI's own track

TechCrunch frames safety as a route to independence for OpenAI from Nvidia, its major investor. The company is building its own safeguards and publicly disclosing the worst incidents it finds. It also runs a separate cybersecurity consortium, the Defense Factory, whose supporters include Anthropic, AWS and Google: many of the names that did not join Nvidia's technology-focused effort. Cybersecurity is also becoming an enterprise offering for OpenAI, from its cyber-oriented Daybreak model to a growing partner network.