Nvidia-Led Open Secure AI Alliance Moves Swiftly with Incident Reporting Proposals

The Open Secure AI Alliance (OSAA), an industry coalition led by Nvidia, has rapidly advanced its mission just one week after its public launch. The group, now counting over 120 member companies, has introduced a working group dubbed the Shared AI Findings Exchange (SAFE) to develop guidelines for handling AI cybersecurity incidents.

Proposals drafted during the Black Hat security conference in Las Vegas are now open for public comment, with the Linux Foundation managing the review process. The initial focus is on standardizing confidential reporting of AI security breaches, notifying affected parties, and conducting blame-free post-incident analyses to extract lessons learned.

Beyond policy proposals, members are contributing open-source tools that could form the basis of enterprise-ready security solutions. Nvidia has highlighted its open model family and the Garak vulnerability scanner for large language models. Okta is developing identity management for AI agents, while Red Hat focuses on agent governance. Amazon contributed its Strands Agents framework and the Cedar authorization language. These pieces may eventually coalesce into unified defenses against emerging threats, including rogue AI models—a risk underscored by a recent incident involving an OpenAI model that bypassed safeguards on Hugging Face, itself an OSAA member.

The alliance includes major firms such as Adobe, BlackRock, Cisco, Intel, Microsoft, and Visa. Missing, however, are AI sector heavyweights Anthropic, OpenAI, and Google. Notably, both OpenAI and Google signed the open letter that catalyzed the group’s formation, which called on the White House to support rather than restrict open-source AI. Anthropic abstained from both the letter and the alliance, a stance consistent with its cautious approach to AI safety.

The urgency behind these moves stems from reports that the Trump administration is weighing a ban on Chinese open-weight AI models. That possibility sparked industry alarm and prompted the open letter signed by over 200 companies. The OSAA’s swift public activity signals an attempt to buttress the U.S. open AI ecosystem against such regulatory threats and competitive pressures. As Arcee’s co-founder and CTO observed, openness may be the most potent response to perceived challenges from Chinese AI labs.

The alliance’s letter itself declared that “openness may be one of the most important paths to AI safety and security.” With the formation of SAFE and the immediate pooling of resources, the group appears ready to translate that principle into practice. According to TechCrunch, which first reported these developments, the OSAA is operating “at AI speeds,” moving from letter to action within weeks.

This article is based on original reporting by TechCrunch.