Robots Atlas>ROBOTS ATLAS
Safety

SAFE

2026ExperimentalPublished: 6 August 2026Updated: 6 August 2026Published
Key innovation
Standardizing how the industry collects and shares findings about security incidents in agentic AI systems - turning individual incidents and near-misses into shared defensive threat intelligence for the whole ecosystem.
Category
Safety
Abstraction level
Pattern
Operation level
ApplicationEvaluation (runtime)
Use cases
Sharing threat intelligence about agentic-AI incidentsConfidential reporting of incidents and near-missesAnalyzing patterns of control failuresPublishing security recommendations for the ecosystemCoordinated disclosure of AI system vulnerabilities

How it works

SAFE defines a confidential reporting pipeline: an organization that detects an incident or near-miss in an agentic system reports the findings in a standardized format. Reports are analyzed for patterns and root causes (which controls failed), and recommendations and warnings for the ecosystem are published on that basis. The framework proposes public notification deadlines: affected organizations as soon as possible, and customers within a defined window (a proposed 72 hours) based on credible evidence, while keeping the source data confidential.

Problem solved

Findings about agentic-AI security incidents (e.g. successful jailbreaks, prompt injection, tool misuse) are today fragmented and rarely shared between organizations, so the same weaknesses are exploited repeatedly. There is no common, confidential exchange channel or shared vocabulary for describing incidents.

Components

Confidential reporting pipelineThe entry point of the exchange.

A standardized, confidential way to submit findings about incidents and near-misses.

Pattern and root-cause analysisTurning reports into knowledge.

Identifying recurring patterns and the control mechanisms that failed.

Recommendations and advisoriesOutput / distribution of defensive knowledge.

Publishing evidence-based recommendations and advisories for the ecosystem.

Notification / disclosure timelinesTiming rules for disclosure.

Public deadlines for notifying affected organizations and customers (a proposed 72 hours for customers).

Official

Implementation

Implementation pitfalls
Dependence on participationHigh

The value of shared knowledge grows with the number of reporters; low adoption limits usefulness.

Fix:Ensure confidentiality, easy reporting and clear benefits for participants.
Confidentiality vs transparency tensionMedium

Disclosing too fast can harm affected organizations; too slowly harms the rest of the ecosystem.

Fix:Clear, staged deadlines (affected parties first, then broader notification) based on credible evidence.

Evolution

2026
SAFE draft published as an RFC (Black Hat USA)
Inflection point

The Open Secure AI Alliance (hosted by the Linux Foundation) announced the SAFE draft as a Request for Comments at Black Hat USA in early August 2026; alliance membership topped 120 organizations.