@OrWeis

Unrelenting entrepreneur Founder of @permit_io, @Rookoutlabs, @Reactful Permissions for the AI Era (https://xitter.cf/t.co/3jbOiNRcAl)

permit.io
Joined September 2013
  • Tweets1K
  • Following155
  • Followers1K
63Media
EN 18286 is the first harmonised standard under the EU AI Act, now in the Official Journal. It covers QMS for high-risk AI. For agent builders, the question it raises is whether you can prove what your agent was allowed to do and who approved it. permit.io/blog/en-18286-qms-…

EN 18286:2026 QMS Standard: What It Means for AI Agent Authorization

EN 18286:2026 is Europe's first harmonised standard under the EU AI Act. It turns Article 17's QMS requirements into auditable processes. Here's what it means for teams shipping AI agents.

permit.io
29
Most agents in prod borrow a shared API key or the user's session. Whatever that key reaches becomes the blast radius, and the audit log says the human did everything.
2
45
more replies
For MCP, OAuth covers whether the client can connect. Whether this agent can call this tool on this resource right now is a separate check. A gateway PEP in front of MCP servers can do it without rewriting each server.
2
20
NIST NCCoE's concept paper frames the same four areas: identification, authorization, delegation, logging. Full post: permit.io/blog/agents-as-pri…

Treat AI Agents as Privileged Identities

Treat AI agents as privileged NHIs: distinct identity, least privilege, attributable delegation, and forensic decision logs, not shared keys. Map to NIST NCCoE's identification · authorization ·...

permit.io
9
NIST already defines what an audit record needs: what, when, where, source, outcome, who (SP 800-53, AU-3). AI agents need more on top of that.
1
1
50
more replies
Teams often skip logging denies. An allow-only trail reads like telemetry. Add denies and it reads like a control.
1
4
All 12 fields, with a sample JSON record: permit.io/blog/ai-agent-audi…

AI Agent Audit Logs: The 12 Fields Auditors Expect

AI agent audit logs need about twelve fields: identity, delegation, tool, resource, policy, decision, HITL, and outcome.

permit.io
6
Congrats to the Oso team on joining Temporal. They pushed authorization forward for years. Oso Cloud teams get 25% extra on every Permit quota for 12 months, hands-on migration help on Enterprise, and this guide: permit.io/blog/migrate-from-…

How to Migrate from Oso to Permit.io

A practical migration guide for teams moving from Oso Cloud to a supported authorization platform. Covers mapping Polar policy patterns to RBAC, ReBAC, and ABAC, step-by-step migration planning, and...

permit.io
1
5
116
Tool traces show what an agent called. They do not prove authorization.
1
1
56
more replies
Emit that at the tool-call PEP, not only at login. A gateway in front of MCP servers is a natural place for it.
1
9
Same bar as human SaaS AuthZ. Moving the interface from UI to prompt doesn't lower what you have to prove. permit.io/blog/human-agent-d…

What Audit Trail Do You Need for Human-to-Agent Delegation?

For human-to-agent delegation, bind human principal, agent ID, tools, data, policy decision, and HITL, not just tool traces, into an attributable audit trail.

permit.io
12
1/ Agents raise the authorization workload. Same security bar. One prompt can turn into dozens of tool calls, and each one needs its own check.
2
43
more replies
5/ Realtime: if you revoke a human and the agent keeps working, that is the incident question. Audit: "the AI did it" will not hold up. Log principal, agent, tool, resource, decision, reason.
1
20
6/ Skip AgentAuthZ 2.0. Extend FGA, put a PEP in front of MCP tool calls. permit.io/blog/ai-era-same-b…

AI Agents Don't Lower the Security Bar. They Raise the Authorization Workload.

AI agents multiply the places where authorization must run. Delegation becomes a first-class part of the authorization shape: the human, the agent, the scope, the resource, conditions. Extend...

permit.io
17
Load more