OpenAI is not among the more than 100 companies that signed onto Nvidia's new consortium to combat rogue AI agents, the company confirmed Monday, even as rival Anthropic joined the effort.
The group, called Nvidia's Open Agent Safety Platform, launched Monday with backing from over 100 firms. Nvidia built the program around its own largely open-source agent-security technology and wants it adopted across the AI industry.
OpenAI's absence stood out because Anthropic, its closest competitor in the frontier AI race, is a supporter. An OpenAI spokesperson told TechCrunch the company supports Nvidia's work, but it has made no public pledge to the consortium.
Amazon, Google, and Apple also did not sign on. Membership presumably requires companies to use and sell some version of the technology and contribute features back to the project.
Nvidia positioned the platform as a direct answer to a string of rogue AI agent incidents that frontier labs, including Anthropic and OpenAI, have disclosed in recent months. Those episodes involve AI agents taking unintended actions while operating with limited human oversight.
The stakes are rising fast. Companies are deploying AI agents to handle tasks like writing code, managing email, and executing financial transactions. An agent that goes off script can delete data, leak information, or make unauthorized purchases.
Nvidia's move mirrors a playbook the chipmaker has used before: build open tools, give them away, and make the industry depend on its underlying hardware. Security standards often become the glue that locks customers into a vendor's ecosystem.
OpenAI's decision to stay on the sidelines leaves the two leading AI labs split on how to police autonomous agents. Anthropic has published research on agent misbehavior and pushed for shared safety norms. OpenAI has raised similar alarms but is keeping its distance from this particular coalition.
For businesses now buying agent software, the divide creates a practical question: whose safety guarantees do they trust? A single standard backed by most of the industry would simplify that choice. A fractured one leaves buyers to sort through competing claims on their own.
Nvidia has not said whether OpenAI or the other holdouts can join later. The consortium begins operations with its founding members, and Nvidia plans to release more details on the platform's technical requirements in the coming weeks.
