An evidence layer for agent authority: Mysten Labs and Google Cloud’s Verifiable Agent Arbiter

An evidence layer for agent authority: Mysten Labs and Google Cloud’s Verifiable Agent Arbiter

by

On 6 October 2026 Mysten Labs, in collaboration with Google Cloud, announced the Verifiable Agent Arbiter (VAA) — an evidence layer being co-developed to let enterprises prove that an AI agent acted within the authority it was granted. The pitch is narrow and pointed: not “trust our agent platform”, but “here is a record a counterparty, auditor or regulator can verify independently of the platform that produced it“.

What happened

VAA separates the record of an agent’s activity from the proof of its integrity:

  • Detailed telemetry — prompts, model outputs, tool calls and policy decisions — stays private, in customer-controlled Google Cloud Storage.
  • Cryptographic proofs of that record are stored on Walrus (Mysten’s data platform) and coordinated on Sui, a Layer 1 blockchain.

The announced capabilities are aimed squarely at disputes and audits:

  • Dispute replay across companies. When two organisations’ agents transact through the open Agent2Agent (A2A) protocol — donated by Google Cloud to the Linux Foundation — both parties can replay a disputed workflow against the same verifiable receipts to establish what was committed, what was delivered, and where it failed.
  • Incident reconstruction. Security teams can rebuild the instructions, policies, decisions, outputs and tool actions around a suspected rogue execution, and distinguish a prompt-injection attack, a policy bypass, a model failure, or a post-incident change to the logs.
  • Durable retention. Archived evidence can be re-proven under future cryptographic standards, including post-quantum, without breaking its link to the original record.
  • End-to-end agentic commerce. VAA connects to the Sui Agent Payments facilitator, which lets agents settle payments for API calls using the x402 protocol; VAA logs the underlying authorisation, spend decisions and service delivery.

Mysten Labs says the driver is the absence of a record that third parties can trust, and cites regulatory logging requirements — including, in its framing, EU AI Act penalties of up to €15 million or 3% of global annual turnover for logging failures from 2027. Initial enterprise deployments are planned ahead of broader availability.

“It brings together records of what an agent was permitted to do, what it did, and what resulted,” said Evan Cheng, co-founder and CEO of Mysten Labs.

Why it matters for KYA

Know Your Agent is often reduced to the moment of authentication: proving which agent is calling. VAA attacks the other half — proving after the fact that the agent stayed inside its grant. That is the part that turns an agent’s identity into something an auditor, a counterparty or a court can act on, and it maps onto the accountability side of KYA: attribution, an audit trail, and evidence of delegated authority.

Three things are worth weighing before treating the announcement as a solved problem:

  • Who defines “authority”? A proof that an agent stayed within its grant is only as good as the grant — VAA anchors the integrity of a record, not the correctness of the policy that produced it.
  • The neutrality claim is the crux. The value rests on a counterparty being able to verify the proof without trusting Mysten, Google Cloud, or the enterprises involved; the design’s soundness and any independent audits are what matter, not the marketing.
  • It is early. “Initial deployments planned ahead of broader availability” means there is no shipped, independently reviewed product to evaluate yet — only an architecture and a set of claims.

For a practitioner, the useful takeaway is directional: as agents begin to negotiate, transact and act across company boundaries, a KYA programme needs an evidence trail that survives leaving the platform that created it. Whether or not VAA wins, “reconstructable, portable proof of what an agent was authorised to do” is the shape of the requirement.

Sources

Know Your Agent (KYA) explains agent identity, verification and accountability. This is an explainer, not legal or compliance advice — see our Sources & methodology. New to KYA? Start with What is Know Your Agent? and the glossary.