The Linux Foundation has put out a public draft, open for comment, of a proposed framework called the Shared AI Findings Exchange (SAFE), with the idea to create a common pool for companies to report security incidents related to AI agents.
Released on August 4 to coincide with the opening of Black Hat conference in Las Vegas, the framework has been drafted by Open Secure AI Alliance, a network consortium launched by NVIDIA on July 27 that now claims more than 120 members.
According to Linux Foundation, the SAFE framework proposes “confidentially collecting and analyzing AI incidents and near misses, notifying affected organizations, identifying recurring control failures, and developing evidence-based recommendations that help reduce systemic risk across the AI ecosystem.”
“Rather than focusing on blame or enforcement, SAFE is designed to promote shared learning while respecting existing legal, contractual, and regulatory obligations,” read a statement from Linux Foundation.
According to the public draft, NVIDIA, Cisco, Crowdstrike, Hugging Face and Red Hat contributed to the proposal, but nothing has been agreed upon till now with no enforcement mechanism or published disclosure timeline for the companies.
“The SAFE guidelines include proposals to confidentially collect and analyze AI incidents and near misses, inform those impacted, identify recurring control failures and publish evidence-based operating recommendations that reduce systemic risk,” read a statement from NVIDIA on Tuesday.
The formation of Open Secure AI Alliance comes after multiple reports emerged of AI models like OpenAI’s GPT-5.6 Sol and Anthropic’s Claude Mythos and Fable 5.0 autonomously hacking other organizations after escaping their sandboxes.
According to the public document, the framework proposes that SAFE should be operated neutrally and its processes are intended to apply equally to both open-weight and closed AI models and systems.
“The proposal also envisions that, where appropriate, SAFE could publish reusable tests, machine-readable policies, detection rules, reference configurations, and incident response guidance, creating a shared catalog of defensive recommendations that evolves alongside emerging AI threats,” read a statement from Linux Foundation.
In the context of the SAFE proposal, alliance members demonstrated several open-source security tools that have been created to help strengthen AI safety and testing. Among other innovations, NVIDIA promoted technologies such as OpenShell, NOOA, NeMo Guardrails, and Garak vulnerability scanner that enhance the safety of AI. The rest of the members introduced innovations ranging from identity security to agent orchestration.
Several companies, such as Amazon, Microsoft, Cisco, CrowdStrike, Okta, Palo Alto Networks, Cloudflare, Visa, Uber, LangChain, and Veeam, also presented new open-source projects aimed at making AI safer. This tool helps software developers secure their AI agents as well as identify vulnerabilities, monitor agents’ work, and control access in case of failure.
Also Read: AISI Flags Claude Mythos and GPT-5.6 Sol for Going Rogue 19 Times in Cyber Tests
