Linux Foundation and Open Secure AI Alliance Propose SAFE Guidelines for Sharing AI Incident Intelligence

On August 4, as the annual Black Hat conference opened in Las Vegas, the Linux Foundation published a request for comments on the Shared AI Findings Exchange (SAFE), a proposed set of guidelines for turning agentic AI security incidents into shared protection for the entire ecosystem.

The SAFE framework, drafted by a working group of the Open Secure AI Alliance, calls for confidentially collecting and analyzing AI incidents and near misses, informing those affected, identifying recurring control failures, and publishing evidence-based operating recommendations that reduce systemic risk.

The alliance now counts more than 120 member organizations, with NVIDIA, Cisco, CrowdStrike, Hugging Face, and Red Hat among those contributing to the initial proposal alongside the Linux Foundation.

The framing reflects a broader shift: an AI agent is not just a model but a system of identity controls, harnesses, guardrails, logs, and evaluation, and securing it requires more than vulnerability scanning. When trusted ecosystems share threat intelligence openly, the alliance argues, collective defense becomes a force multiplier.

NVIDIA paired the announcement with a stack of open-source security contributions, including the NOOA research harness for testing and auditing agent behavior, the OpenShell runtime that enforces security and privacy controls at the agent level, verified agent skills that are scanned for risks and cryptographically signed, and tools such as NeMo Guardrails, NeMo Anonymizer, and the Garak LLM vulnerability scanner.

Other members added tools across the defensive stack: Okta is building reference implementations for agent identity and access, Palo Alto Networks open-sourced Agent Guard and Agent Watch from its Idira platform, and Red Hat launched asago, a project that maps governance requirements such as NIST, OWASP, and the EU AI Act to runtime agent permissions. Amazon also joined the alliance on the same day, contributing its open-source Strands Agents toolkit.

The signal is that the industry is starting to treat agent security as infrastructure on par with model capability. SAFE is still in the comment phase, so the questions to watch are how the RFC feedback shapes a formal standard — and whether the alliance's tools actually work together after the Black Hat buzz fades.