Skip to content
NewA calmer, faster ZeroHour

ZeroGate: Trust-Preserving Fast Paths for Governed AI Agent Runtimes

arXiv cs.CRby Zexun Wang ()1 min read2 · 1 read
AI summary · grok-4.7Summaries by Vani AI

ZeroGate proposes signed short-lived ActionPasses so AI agent runtimes recheck authorization before local admission.

ZeroGate separates exact-action approval from local admission: an issuer signs a short-lived ActionPass, and a trusted runtime adapter reconstructs the action before a local gate checks its binding and consumes a nonce. A SQLite transaction atomically couples nonce consumption, quota updates, and an admission receipt, under a conditional decision-preservation claim. Across 4,800 Azure Blob attempts, prepared worker-admission-to-dispatch p95 was 9.802–11.374 ms versus 25.018–334 ms synchronously, but mean complete lifecycle was longer for the prepared path. The authors offer a revalidation contract and cost comparison, not a new cryptographic primitive or a net speedup.

Key facts
  • An issuer signs a short-lived ActionPass checked again at local admission.
  • A SQLite transaction atomically consumes the nonce, updates quota, and records admission.
  • Prepared dispatch p95 was about 10–11 ms versus up to 334 ms synchronous.
  • Prepared execution did not reduce mean complete lifecycle latency.
  • Local admission alone does not guarantee freshness or exactly-once remote effects.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

From the original article

1 min read at arxiv.org

Moving authorization earlier can shorten an agent's dispatch boundary without removing authorization work. It can also admit an action whose payload, authority, or relevant state has changed.

arXiv cs.CR

By Zexun Wang · arxiv.org

Read the full article at arXiv cs.CR

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.