A group of over 120 tech organizations have published plans for a new information-sharing initiative focused on agentic AI threats.

Members of NVIDIA’s Open Secure AI Alliance announced details of the Shared AI Findings Exchange (SAFE) on August 4.

The Linux Foundation explained in a blog post on the same day that organizations today often investigate AI security incidents internally, meaning potentially valuable information remains locked away.

“There is no broadly adopted community framework for confidentially sharing AI operational failures, identifying recurring control failures, and translating those lessons into reusable defensive guidance across the ecosystem,” it continued.

“The SAFE proposal is intended to help the industry collaborate in moving that conversation forward.”

Read more on AI threats: AI Agents Now the Enterprises Fastest Growing Exposed Attack Surface

The idea is to focus on shared learning rather than blame or enforcement, while respecting existing legal, contractual, and regulatory obligations, the Linux Foundation said.

Backers have published an open RFP to ensure the broader community can develop the SAFE guidelines together.

However, a draft proposal is built around the following principles:

• Confidential reporting of AI security incidents and near misses
• Timely notification to any affected organizations
• Collaborative analysis focused on shared learning
• Structured reviews spanning the complete AI operating stack, including models, safeguards, tools, runtime environments, monitoring, human operations, and supply chain dependencies
• Evidence-based operational guidance for organizations to implement and verify
• Independent governance representing stakeholders from across the AI ecosystem, with no single vendor able to control the group’s findings

The group’s processes are designed to apply equally to open and proprietary AI 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,” the blog continued.

Agents Pose Mounting Risks

Although the Linux Foundation does not mention AI agents specifically in its post, an NVIDIA blog announcing SAFE explicitly refers to the initiative as one designed to “transform agentic cybersecurity incidents into better protection.”

This is certainly the area where AI technology seems to have the greatest potential for unintended consequences.

Both OpenAI and Anthropic models have been caught attacking real-world targets during recent testing.

Just this week, the UK’s AI Security Institute (AISI) reported that models from both companies engaged in “sustained, potentially harmful activity” targeting real people and organizations.

Jacob Krell, senior director for secure AI solutions and cybersecurity at Suzu Labs, argued that SAFE is needed because existing vulnerability disclosure mechanisms don’t fit the AI agent model.

“When a model finds and uses access it shouldn't have reached, there's no patch to issue and no vulnerability identifier to publish. Agent failures are often behavioral and non-deterministic, with no signature to match and no fix to deploy,” he added.

“The highest value will come from near misses. Breaches make headlines, but an agent that probes a boundary and fails never gets published. That behavioral pattern is exactly the intelligence other organizations running similar systems need, and SAFE creates the channel for sharing what would otherwise stay invisible.”

Jeremiah Fowler, researcher for Black Hills Information Security, also welcomed the news.

“The more organizations that contribute real-world evidence, the more effectively we can identify emerging attack patterns, common vulnerabilities, and evolving threats while reducing duplicated defensive efforts that waste valuable time and resources,” he argued.

“As AI becomes increasingly integrated into everyday life, business operations, and critical infrastructure, developing standardized security guidance now is a proactive investment rather than waiting to retrofit a defense after incidents occur.”