What exactly did GMO and VESS Labs test?

They connected a buyer-side AI agent and a seller-side AI agent developed independently by different companies. The test covered mutual identity and authority verification, human purchase approval, ordering, and mock payment.

Why can't companies just check an AI agent's credentials themselves?

In a cross-company setting, a firm would have to ask the other company's systems individually to confirm which business the agent belongs to, how far its trading authority goes, and who approved the deal. A shared trust framework is meant to avoid that per-inquiry checking.

Was real money used in the trial?

No. The proof-of-concept ran through to a mock payment, not a live settlement.

Large Language ModelsAI Regulation & PolicyAI Business & IndustryITmedia AI＋8h agoSep 25, 2026, 13:00 JST

GMO and VESS Labs prove cross-company AI agent trade trust

Send on LINE

3 Key Points

What happened
GMOグローバルサイン・ホールディングス and VESS Labs ran a joint 信頼基盤 proof-of-concept from 2026年4月～9月, linking buyer-side and seller-side AIエージェント built independently by different firms through identity and authority checks, human approval, ordering, and mock payment.
Why it matters
Cross-company AI deals today would force each firm to query the other's systems to confirm who an agent belongs to and what it may do, so a shared trust layer could remove that step-by-step checking.
What to watch
The test covered mock payment and human sign-off, so real adoption hinges on whether production payment rails and liability rules follow. Watch whether the two firms extend the 信頼基盤 work beyond the April–September 2026 window.

WHO IT HITSThis lands hardest on corporate procurement and payment operations teams that would otherwise have to manually verify the identity and spending authority of an outside company's AI agent before letting it place orders.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Get news like this every morning, free

Context & Analysis

The proof-of-concept sits at the point where AI agents stop being an in-house tool and start acting as counterparties to other firms. Inside one company, the relationship between an agent, its owner, and its user can be managed in advance. Once agents from separate companies connect directly to trade, that certainty disappears.

The two partners addressed this by having independently developed buyer-side and seller-side agents verify each other's identity and authority, then routing the transaction through human purchase approval, ordering, and a mock payment. The trial ran from April to September 2026.

The open question is whether a shared trust layer can carry real settlements. Because the test stopped at mock payment, the outcome hinges on whether production payment rails and clear responsibility rules follow — a question most relevant to the procurement and finance teams that would authorize an outside agent to spend on their behalf.

FAQ

What exactly did GMO and VESS Labs test?: They connected a buyer-side AI agent and a seller-side AI agent developed independently by different companies. The test covered mutual identity and authority verification, human purchase approval, ordering, and mock payment.
Why can't companies just check an AI agent's credentials themselves?: In a cross-company setting, a firm would have to ask the other company's systems individually to confirm which business the agent belongs to, how far its trading authority goes, and who approved the deal. A shared trust framework is meant to avoid that per-inquiry checking.
Was real money used in the trial?: No. The proof-of-concept ran through to a mock payment, not a live settlement.

ITmedia AI＋Read Original Article

Send on LINE

Get the latest Large Language Models news every morning

For example, today's edition would include:

Okta, CrowdStrike, Google Cloud form Blueprint Alliance for AI agent securityYahoo Finance AI · 2h ago
White House tells OpenAI, Anthropic: hold models from UK testersTHE DECODER · 2h ago
Ruby on Rails creator David Heinemeier Hansson quits hand-codingTHE DECODER · 2h ago

AI-summarized, only the topics you pick: one digest a day via Email, LINE, or Slack.

Subscribe free →

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.