120 companies want a rule for telling you when their AI agent broke into something
The Open Secure AI Alliance, a group hosted by the Linux Foundation whose backers include Nvidia, Cisco and CrowdStrike, published a proposed framework on 11 August called SAFE — Shared AI Findings Exchange — for reporting security incidents caused by AI agents. It covers agents gaining unauthorised access to third-party systems, disclosing confidential information, or continuing to probe a live target after the operator suspects something has gone wrong, and it covers near misses as well as actual harm. Members would preserve prompts, agent traces, tool calls, identities, permissions and credentials as evidence; notify affected organisations as soon as possible and demonstrably exposed customers within 72 hours; file a confidential report within four business days and a public one, with an analysis of what failed, within 30 days. It is voluntary, and for now it is only a document: there is no way to submit a report and no implementation timeline, just a GitHub repository collecting feedback.
Related stories
Nevada just cleared up to 8,000 driverless taxis for Las Vegas — Tesla's ceiling went from 10 to 5,000 in five weeks2026-08-21DeepSeek's cheap workhorse can now see — and on agent tasks that need eyes it says it is close to Anthropic's best
2026-08-21OpenAI has started a team whose stated worry is that a company like OpenAI ends up with too much power
2026-08-21Given four hours and a GPU to improve the way AI is trained, the best agent scored 0.25 out of 1 — and most never tried
2026-08-21ChatGPT can now read your iMessages and send them — and the setting that lets it skip asking is the one OpenAI warns about
2026-08-21