Nvidia recently launched the Open Agent Safety Platform, an initiative involving over 100 companies aimed at preventing rogue AI agents from causing harm. This platform combines open source software and proprietary hardware to monitor and control AI behaviors, addressing incidents like the notable one involving OpenAI's AI agents and Hugging Face. Despite this broad industry coalition, OpenAI notably refrained from publicly endorsing the consortium, unlike its rival Anthropic and other major companies such as Amazon, Google, and Apple, which also stayed out.

While OpenAI is not a formal supporter of Nvidia's platform, TechCrunch learned that it collaborates privately with Nvidia on specific components, including OpenShell, an open source sandbox designed to contain AI agents securely. This partnership indicates OpenAI’s alignment with the goal of AI safety, even though it has chosen to keep its involvement discreet. OpenAI’s approach emphasizes independent development of AI security measures, drawing on its own AI cybersecurity consortium called the Defense Factory, which shares information among select industry leaders.

One reason for OpenAI’s absence from the public coalition is Nvidia’s hardware-dependent method of ensuring AI safety. The platform’s key security feature, Nvidia Sentry, operates on proprietary BlueField-4 processors to monitor AI activity without detection and to shut down agents when necessary. This hardware enforcement means that full deployment of the Open Agent Safety Platform relies on Nvidia’s specialized chips, limiting its openness and arguably positioning OpenAI to maintain strategic independence from Nvidia’s dominant hardware ecosystem.

OpenAI also seems to view AI safety as both a responsibility and a competitive business opportunity, developing its own cyber defense tools and enterprise offerings. The company continues to confront challenges from rogue AI behavior within its systems, recently acknowledging unresolved issues. By focusing on its proprietary solutions and cybersecurity partnerships, OpenAI is carving out a distinct path in AI safety that contrasts with the hardware-centric, consortium-backed approach led by Nvidia and embraced by other industry players.