Opens in a new tab

Advanced AI Society Releases Proof-of-Control v1.0 Draft for Open AI Agent Verification

  • Home
  • Blog
  • Advanced AI Society Releases Proof-of-Control v1.0 Draft for Open AI Agent Verification
AI agent Head inside a glowing proof of control verification chamber

Advanced AI Society has released the Proof-of-Control v1.0 working draft for public comment. The official Proof-of-Control page says it is an industry-led open standard for evidence that an AI agent stayed within the controls it was given. The draft, hosted by LF Decentralized Trust, is an open standard rather than a finished compliance requirement.

The official Proof-of-Control page says the draft is open for live commentary through November 15, 2026. Advanced AI Society’s announcement materials describe the proposal as a way to provide evidence that an agent stayed within the controls it was given, without asking a business to rely only on an operator’s account or a vendor’s assertion.

What Proof-of-Control Requires

The Advanced AI Society announcement states that an agent’s authority should be defined before it acts, that runtime evidence should show whether it remained within that authority, and that the evidence should remain auditable across the agent lifecycle.

Proposal element Source Question for an IT manager
Predefined authority Ken Huang, Co-Chair of Proof-of-Control, in the Advanced AI Society announcement Can the team identify what the agent is allowed to do before deployment?
Runtime evidence Ken Huang in the announcement Do runtime records show the actions the agent actually performed?
Lifecycle auditability Ken Huang in the announcement Can an authorized reviewer examine the evidence over the agent’s full lifecycle?

The official project page says anyone can verify the evidence without trusting the operator or an auditor, and any change made after it is recorded can be detected. It also describes Proof-of-Control as an industry-led open standard maintained by the Advanced AI Society’s Open Verification Lab and hosted as a lab at LF Decentralized Trust.

Why IT Teams Need Action-Level Evidence

AI agents can act faster than annual or quarterly reviews can capture. Tricia Wang, Co-Founder and CEO of Advanced AI Society, said in the announcement that “Safety checks and traditional audits are necessary,” but added that an annual or quarterly approval does not cover actions taken in milliseconds.

The project’s approach asks IT teams to keep evidence tied to individual actions and make that evidence available to a reviewer other than the vendor or operator.

That concern extends to policy proposals for AI accountability. A October 7, 2026 discussion draft of the CLAIM Act would hold developers liable in specified circumstances when AI systems engage in conduct that would be unlawful if committed by a person. The draft also describes transparency, independent verification and incident reporting requirements under the related FRONTIER Act.

Who Is Behind the Proposal

Advanced AI Society says the draft was co-designed with more than 80 global security leaders across enterprise sectors and built from verifiable AI technologies developed by member companies. Advanced AI Society announced its membership in the Linux Foundation and LF Decentralized Trust, and Proof-of-Control is hosted as a lab at LF Decentralized Trust.

  • Tricia Wang, Co-Founder and CEO of Advanced AI Society
  • Ken Huang, Co-Chair of Proof-of-Control
  • Daniela Barbossa, General Manager, Decentralized Technologies, Linux Foundation, and Executive Director, LF Decentralized Trust
  • Dr. Hart Montgomery, CTO of LF Decentralized Trust
  • Charles Iheagwara, Global Head, AI Cybersecurity at AstraZeneca
  • Bruce Schneier, Cryptographer and Technical Advisor
  • J. Christopher Giancarlo, former Chairman of the U.S. Commodity Futures Trading Commission and Senior Advisor at Advanced AI Society
  • Clay Shirky, Vice Provost for AI and Technology in Education at New York University
  • Noah Ringler, former AI Policy Lead at the U.S. Department of Homeland Security and Senior Advisor at Advanced AI Society
  • Michael Casey, Sheila Warren and Bettina Warburg, Advanced AI Society Co-Founders
  • Jim Schwoebel, Director of Research at Advanced AI Society and CEO of Quome

Ken Huang said in the announcement that “An agent’s runtime log is just a claim until anyone can verify it for themselves.” The proposal uses precedents including AnonCreds and Trust Over IP, according to the announcement.

How Should an IT Team Assess Proof-of-Control?

Proof-of-Control remains a working draft and has not been presented as a finished implementation requirement.

  • Agent inventory: List every autonomous or AI-assisted process that can change business systems, including email, websites, servers, records and financial workflows.
  • Documented permissions: Record the authority granted to each agent, including the actions it can take and the systems it can reach.
  • Action-level records: Check whether runtime logs capture what the agent did rather than only recording that it ran.
  • Independent review access: Determine whether someone other than the vendor or operator can examine the records.
  • Retention: Check whether evidence remains available throughout the agent’s lifecycle and for incident review.
  • Escalation: Identify who reviews an unexpected change and how the review is recorded.

Questions to Ask a Vendor

  • Where is an agent’s authority documented before deployment?
  • What does the runtime log contain for each action?
  • Can an authorized party other than the vendor inspect the evidence?
  • How long are records retained?
  • How are unexpected actions surfaced for review?
  • What controls can produce evidence without exposing private or sensitive business data?

What to Do Now

  1. List every autonomous or AI-assisted process that can change business systems.
  2. Document each agent’s permissions, accessible systems and responsible approver.
  3. Review available runtime records for individual actions.
  4. Identify who can examine those records outside the vendor or operator.
  5. Ask each vendor for its evidence retention, incident review and escalation procedures.

The draft can be reviewed and commented on through November 15, 2026, according to the official Proof-of-Control page. Advanced AI Society and LF Decentralized Trust also scheduled a public launch event for September 23, 2026, according to the organization’s announcement.

FAQ

What Is the Proof-of-Control Standard for AI Agents?

Proof-of-Control is an industry-led open standard for evidence that an AI agent stayed within the controls it was given. The v1.0 working draft calls for authority to be defined before action, runtime evidence of the agent’s actions, and auditability across the agent lifecycle.

Who Is Hosting Proof-of-Control?

LF Decentralized Trust is hosting Proof-of-Control as a lab. The project is maintained by the Advanced AI Society’s Open Verification Lab.

How Can the Public Comment on the v1.0 Draft?

The official project page says Proof-of-Control v1.0 is open for live commentary through November 15, 2026 and provides a link for commenting on the Google Doc.

BizScoreAI

BizScoreAI, which includes the free listing check

BizScoreAI scores how visible a business is to AI search and shows what its listing looks like to the engines people ask. Open the free listing check.

This article reports on publicly available information from the sources listed below. We have no reason to believe it is inaccurate, but we have not independently verified every claim, and details may change as the story develops.

Sources

← All Articles