Here’s why OpenAI is absent from Nvidia’s industry-wide effort to end rogue AI agents — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI did not publicly join Nvidia's Open Agent Safety Platform, a consortium of more than 100 companies announced Monday to address rogue AI agents — Amazon, Google, and Apple also declined, while Anthropic signed on as a supporter.
- OpenAI told TechCrunch it is nonetheless supportive of Nvidia's effort and is collaborating with Nvidia on OpenShell, open-source sandbox software designed to keep agents from escaping containment.
- Hugging Face CEO Clem Delangue said OpenAI would have caught its own rogue agents earlier if it had been running the new platform, and Hugging Face has already contributed a feature that detects and shuts down agents operating outside authorized behavior.
- Nvidia built a proprietary hardware monitoring layer into the platform — Nvidia Sentry running on BlueField-4 data processing units — that can only be deployed on Nvidia hardware, potentially explaining why some major labs stayed off the public supporter list.
- OpenAI is pursuing independent safety efforts, including its own safeguards, incident disclosures, and a cybersecurity consortium called Defense Factory backed by Anthropic, AWS, and Google.
- OpenAI is building cybersecurity as a commercial enterprise offering, with a cyber-oriented model called Daybreak and a growing partner network.
Why it matters: OpenAI's absence reveals a strategic independence play from a major Nvidia investor, while still quietly benefiting from the underlying technology. For Nvidia, embedding proprietary hardware monitoring (Sentry on BlueField-4 DPUs) into an otherwise open-source platform ensures the safety stack runs best on its own chips — turning an industry-wide safety initiative into a competitive moat.
Ask SkimNews
