Nvidia-Backed OSAA Races Ahead with New Working Group and Open-Source Tech

A week after its formation, the Open Secure AI Alliance (OSAA), led by Nvidia, has already made significant progress. The industry group, which now boasts over 120 companies, has established a working group called the Shared AI Findings Exchange (SAFE). This move demonstrates the alliance's commitment to collaboration and open-source technology.

The SAFE working group aims to develop guidelines for confidentially reporting AI cybersecurity incidents, alerting those affected, and conducting blame-free analysis. These proposals are now open for public comment, with The Linux Foundation managing the process.

Members of OSAA are also contributing their open-source technologies to the effort. Nvidia has shared its family of open models, as well as an LLM vulnerability scanner called Garak. Other companies, such as Okta and Red Hat, are working on agent identity tech and agent governance, respectively. Amazon has contributed tools like Strands Agents and Cedar.

The group's rapid progress is a testament to the industry's willingness to work together in the face of potential threats. Notably, OpenAI and Google, which initially signed an open letter urging support for open-source AI efforts, have yet to join the alliance. However, their presence at Black Hat this week may indicate a shift towards collaboration.