- Ian Livingstone of Keycard said the first problem is old but worse with agents: how to treat traffic that arrives unregistered and unknown.
- Dick Hardt argued for interoperable standards, and another speaker asked whether the industry will favour large players or build federated protocols.
Ian Livingstone, CEO of Keycard, said on the AI Engineer show that the first problem is an old one that agents make worse. It is how to treat traffic that arrives unregistered and unknown. Two panel speakers argued for open standards, not closed systems.
What was said
A shopper who has not logged in arrives at eBay, and eBay does not know who they are. Livingstone said captchas were added to stop spam, because only a human could solve them. Agents have now broken them. In his words, "this has been a traditional problem on the web."
He then described the Vercel block in Spain, two to three years ago. Vercel, a hosting company, shared the same IP addresses across every site it hosted. One site carried Premier League content. Spain wanted to stop it showing illegally, so it blocked all of Vercel.
Agents have the same problem, he said, only "so much worse." The question is "how do you have unregistered traffic you've never met before show up to your front door", and then decide whether to admit it. He wants shared rails so a site can choose which agents to trust.
Dick Hardt, co-creator of OAuth, said many people are building proprietary, closed systems. He likes the personal agent protocol idea but said it does not go far enough. Open standards let you plug in parts from different places. Otherwise, he said, the agent's vendors decide the guardrails.
Another speaker, not clearly identified, said federated means each site decides what it allows, on shared standards. That speaker asked if the industry will pick things that "bias to the large players and give them like concrete block doors". Or will it build federated protocols like OIDC (a sign-in standard) and TLS (the standard behind secure connections)?
Why it matters
Our reading: the Vercel case shows what happens when a blocker cannot tell one tenant from another. A shared address left Spain one crude tool: block everyone behind it. Agents with no identity leave sites in a similar spot.
A federated scheme lets each site choose whom it trusts using common rules. In a closed scheme, a vendor sets the terms. Buyers can ask whether an agent identity product works with other vendors' parts.
The other side
The excerpts do not say how a site would come to trust an unknown agent under an open standard. The large-player point was posed as a question, not a settled view.
The Vercel parallel is loose. Spain acted over illegal content, not bot abuse. Keycard moderated the panel, and Livingstone is its CEO. Hardt did not reject the personal agent protocol, so the choice is not simply open against closed.
Written by the WebPulse Newsroom with AI assistance, and checked by our editorial review: every quotation was verified against the recording's transcript. How we use AI.
The conversation this talking point comes from
- AI Engineer: An AI Agent Got Its User Banned From Resy — Keycard, JPMorgan Chase & AAuth (2026-10-17)





