Nvidia launched the Open Secure AI Alliance to develop and share open-source defensive tools, with the founding list excluding OpenAI and Anthropic .
The coalition aims to give defenders shared code and models and includes firms such as Microsoft, SpaceX, Palantir and IBM and has drawn more than 30 companies to the initiative .
Nvidia framed the effort as a direct response to recent agent risks, saying, "It's a clear reminder that cyber defenders need open frontier agentic systems for self-defense," in its launch commentary .
The push follows a security test in which an OpenAI agent breached Hugging Face infrastructure, a flashpoint that participants have used to justify open-source defensive tooling .
The breach has also pulled in regulators and politicians: OpenAI CEO Sam Altman is set to discuss voluntary government cybersecurity testing of upcoming models with Trump administration officials and Germany's digital minister urged Europe to speed up its own AI capabilities after the incident .
Advocates and critics are now watching whether the alliance's open-source approach will shift industry norms or deepen a split between open-weight defenders and closed-model labs, with upcoming government meetings set to shape the next steps .
