ACTIVE  ·  BUILDING  ·  v2026.11.20 BUILD 2026-11-20  ·  JL:IOTA:001
No. 066 · 2026-06-17

The Permission Wall

Sixty million dollars just went to a company whose entire product is proving an agent is allowed to do what it's about to do.

DISPATCH  ·  LOGGED WITH MAI

Arcade raised $60 million last week. Their product answers one question: can this agent, acting for this user, perform this action on this system? That’s it. Authorization for AI agents in production.

Same week, Ping Identity shipped runtime identity for agents across three clouds. Aembit wired agent governance into Microsoft Copilot Studio. KPMG and Microsoft announced Agent 365 for managing agents at enterprise scale. Four companies, four products, one problem.

Agents stay stuck in pilot when security teams cannot approve overprivileged service account access without an audit trail. One hallucination with admin credentials is a data breach. Everyone knows this. Most teams are deploying anyway and hoping nobody asks too hard.

The organizations moving agents into production solved this early. Before the pilot started. They built identity, scoped permissions, and per-action authorization into the infrastructure from day one. The ones still treating agents like chatbots with extra steps are writing the same pilot extension memo for the third quarter in a row.

The model was ready months ago. The question is whether anyone can prove they’re allowed.

LOGGED WITH MAI  ·  2026-06-17  ·  No. 066
← All Dispatches