OpenAI is missing from NVIDIA's safety coalition, but still works on its tech
OpenAI told TechCrunch it supports NVIDIA's work and collaborates on agent security, despite not appearing as a public supporter of the new coalition.
The absent name does not mean no collaboration
OpenAI is not listed as a public supporter of NVIDIA’s new agent safety coalition, according to TechCrunch’s September 29 report. Anthropic is among the supporters. An OpenAI spokesperson nevertheless told the publication that the company supports NVIDIA’s work, and the report says the companies collaborate on agent security, including OpenShell.
That makes this a story about the scope of participation, not proof that OpenAI has rejected agent safeguards or ended cooperation with NVIDIA.
Why the platform design matters
NVIDIA’s September 28 announcement pairs OpenShell, an open source runtime boundary for agent actions, with Sentry, a monitoring system running on BlueField-4 processors. NVIDIA says OpenShell can be extended to third-party compute platforms, while the announced Sentry design uses its own hardware.
The distinction matters when evaluating a shared safety initiative: software portability and adoption of a particular hardware system are different commitments. NVIDIA’s performance and containment claims remain vendor claims; this report is not an independent security evaluation.
What has not been explained
The reviewed sources do not establish one definitive reason why OpenAI has not joined as a public supporter. TechCrunch discusses platform dependence and competing initiatives as context. Those possibilities should not be mistaken for a confirmed explanation from OpenAI.