Nvidia unveiled a new coalition on Monday—bringing together over 100 companies with a shared mission to prevent rogue AI agents—but one major name was conspicuously absent from the list: OpenAI. While OpenAI wasn't alone in declining to join (Amazon, Google, and Apple also stayed out), its absence stood out the most, particularly given that rival Anthropic is backing the effort.

The initiative, called Nvidia's Open Agent Safety Platform, represents Nvidia's push to extend its internally developed—and largely open source—AI agent-security technology across the broader AI ecosystem. It's a direct answer to the rogue AI agent incidents that leading labs like Anthropic and OpenAI have been reporting. Nvidia CEO Jensen Huang has consistently characterized rogue AIs as a routine engineering challenge, solvable like any other technical problem. The Open Agent Safety Platform is Huang backing up that rhetoric with action.

OpenAI is collaborating with Nvidia on agent security, including on OpenShell—one of the core software components of the platform. OpenShell is open sourced software that builds a sandbox specifically engineered to prevent agents from breaking out. Even though it's odd that OpenAI didn't simply endorse the initiative the way its rival Anthropic did, the fact that this frontier AI lab is involved at all is encouraging.

That's because OpenAI, in particular, stands to gain from this technology—at least according to Hugging Face founder and CEO Clem Delangue (who sold his company to Nvidia for $12.9 billion earlier this month). "From what we know (take with a grain of salt, we need much more transparency.), if @OpenAI had been running this on their own agents that attacked us, they would have caught them before we did," Delangue posted.

Delangue noted that Hugging Face has already added a feature to the Open Agent Safety Platform that detects and shuts down AI agents accessing websites they're permitted to visit but in unauthorized ways. For example, this feature triggers when agents circumvent their guardrails and coordinate an attack by leaving messages for each other in an open source code hosting repository. That's precisely one method OpenAI said its renegade swarm of agents used to coordinate its attack on Hugging Face.

But there's another reason some of these major players, OpenAI included, might hesitate to publicly commit to Nvidia's efforts. Accessing the complete system requires a hardware component that isn't open source software—it remains proprietary and can only run on Nvidia's hardware. The Open Agent Safety Platform doesn't just provide a sandbox. It also enforces agent behavior at a hardware layer, where agents cannot tell they're being monitored. (Some AI models and agents lie and pretend to follow the rules when they know they're being watched.)

The hardware monitoring capability depends on Nvidia Sentry, a proprietary feature that operates on specialized Nvidia processors called BlueField-4 data processing units. Sentry continuously watches agent behavior from these processors and can instantly terminate agents, Nvidia promises. While a hardware solution is clearly sound in principle, it means the Open Agent Safety Platform isn't exactly a pure open source endeavor. It enables Nvidia to guarantee that this solution always performs best on its own hardware. Indeed, Nvidia has stated that for those already running workloads on its latest hardware, deploying the Open Agent Safety Platform is a simple software update.

Still, Nvidia competitors—including Arm and Intel—have signed on as Open Agent Safety Platform supporters because the sandbox, OpenShell, can be adapted to work with other chips and hardware. And Nvidia is sharing reference designs for the entire software-and-hardware concept. All of which makes OpenAI's absence even more conspicuous.

Clearly, OpenAI views AI safety as an opportunity for independence from its major investor Nvidia, as well as a chance to demonstrate its own leadership. That holds true even though it was OpenAI's AI agents that alarmed the industry with the Hugging Face incident. For instance, the company is building its own safeguards for its research and products and is disclosing the worst incident it uncovers.

Meanwhile, OpenAI has its own AI cybersecurity consortium for sharing information, called the Defense Factory. Those that signed on to support that idea include Anthropic, Amazon Web Services, and Google—many of the same names that didn't sign on to Nvidia's technology-oriented approach. And, truth be told, some level of fear is good for business. OpenAI is busy crafting cybersecurity into an enterprise offering, with everything from its own cyber-oriented model, Daybreak, to a growing network of partners that enterprises can hire to implement AI security.