OpenAI has chosen not to join Nvidia's new Open Agent Safety Platform, which aims to secure AI agents using hardware-level controls. While other major AI firms have joined the initiative, OpenAI is building its own independent security systems. This move highlights a strategic split between the leading AI developer and its primary hardware supplier as both companies expand into enterprise security services.
Nvidia recently launched the Open Agent Safety Platform, a new consortium designed to standardize security for AI agents. The platform aims to embed safeguards directly into software and hardware, using dedicated data processing units to monitor and control AI behavior in real-time. Over 100 firms have joined the initiative to help create these security standards, including industry peers like Anthropic, Intel, and Arm.
OpenAI, however, is a notable absentee from this group. The company is choosing to maintain independence in its safety protocols rather than adopting the proprietary hardware-based security layer proposed by Nvidia. Instead of relying on a hardware-centric security model, OpenAI is building its own cybersecurity business. This effort involves the development of proprietary research, such as its specialized Daybreak model, and the formation of the Defense Factory consortium. These projects are positioned as independent alternatives to the security solutions offered by hardware manufacturers.
This decision is a significant shift in the AI industry. Software-focused AI developers are increasingly wary of hardware lock-in, where their product security becomes dependent on a specific hardware provider’s technology. By building its own security infrastructure, OpenAI is trying to ensure it has full control over its software and security standards, regardless of the hardware being used.
For investors, this split shows how the relationship between hardware suppliers and AI developers is evolving. Nvidia, which dominates the AI chip market, is now entering the software and security layer of the business. Meanwhile, AI labs like OpenAI are expanding their own capabilities to offer security as a core product for enterprise clients. This means the two companies are increasingly competing in the same space. Future developments to watch include how enterprise clients balance their choice between Nvidia-secured hardware environments and independent, standalone AI security models developed by firms like OpenAI.