Proof retweeted
๐จ Ep. 15 of Agentic Commerce:ย Giving Agents an Identity So They Can Pay
@sytaylor & @bamazizimesh are joined by:
๐ @patk, Founder & CEO,ย @proof
To discuss:
๐ซ Deepfakes, cybersecurity threats and identity challenges in agentic commerce
๐งโโ๏ธ Why KYA and KYC alone cannot establish legal authorization
๐ Comparing Google, Shopify, Visa and Mastercard's agentic commerce protocols
๐ชช How x401 uses verifiable credentials to authorize agent transactions
๐ Why financial institutions block AI agents and need secure access
๐ Intent mandates, step-up authentication and legally enforceable transaction evidence
๐ธ Stablecoin micropayments and combining identity with payment authorization
๐๏ธ FinCEN guidance, trusted credential issuers and identity verification liability
๐ Delegating agent spending authority through signed mandates and revocable credentials
โ๏ธ Implementing x401 through Cloudflare and transaction-specific authorization policies
***
Timestamps:
00:00 Introduction
04:00 Deepfakes, cybersecurity threats and identity challenges in agentic commerce
04:59 Why KYA and KYC alone cannot establish legal authorization
06:23 Comparing Google, Shopify, Visa and Mastercard's agentic commerce protocols
10:44 How x401 uses verifiable credentials to authorize agent transactions
15:07 Why financial institutions block AI agents and need secure access
17:06 Intent mandates, step-up authentication and legally enforceable transaction evidence
19:39 Stablecoin micropayments and combining identity with payment authorization
23:36 FinCEN guidance, trusted credential issuers and identity verification liability
28:09 Delegating agent spending authority through signed mandates and revocable credentials
30:16 Implementing x401 through Cloudflare and transaction-specific authorization policies
***
๐ ๐๐ฆ๐ข๐ณ๐ค๐ฉ '๐๐ฐ๐ฌ๐ฆ๐ฏ๐ช๐ป๐ฆ๐ฅ ๐๐ฐ๐ฅ๐ค๐ข๐ด๐ต' ๐๐ฏ ๐ ๐ฐ๐ถ๐๐ถ๐ฃ๐ฆ. ๐๐ฑ๐ฑ๐ญ๐ฆ, ๐๐ฑ๐ฐ๐ต๐ช๐ง๐บ ๐ฐ๐ณ ๐ข๐ฏ๐บ ๐๐ฐ๐ฅ๐ค๐ข๐ด๐ต ๐๐ญ๐ข๐บ๐ฆ๐ณ! ๐
Weโre live in San Diego for @FintechNerdCon! ๐ค
Join @patk for the main stage session tomorrow at 9:50am PT for a discussion with fraud leaders from @SocureID and @sardine: who is winning the fraud war?
And catch us in the Builder Zone at 11:00am PT as the Proof team takes you through x401 hands-on.ย
๐ Session details (and how you can connect with us during the show): hubs.la/Q04FqWvt0
3
108
Update: our MCP is now in @ChatGPT ๐๏ธ
On Oct. 15 we launched x401 so any business could ask an AI agent a simple question before acting on it: who sent you, and did they say yes? Developers could wire our MCP server into ChatGPT by hand.
Today that step is gone.
Anyone using ChatGPT can add the Proof app, and when the agent hits a site that requires x401, the person gets the exact terms on their phone and approves with their face or fingerprint. The business verifies the signature itself. Change one word and it fails.
ChatGPT gets to do more for you. You stay in charge.
๐ Add Proof to ChatGPT: hubs.la/Q04FCLxZ0
Most of us still approve a wire by typing in a six-digit code from a text, usually after digging around for our phone. During #Money2020, @patk told @IanKhanFuturist that banks donโt trust that code anymore - and that we're building solutions like Certify to replace it.
With Certify, the person approving a wire or a beneficiary change signs that exact request with a cryptographic key tied to their verified identity. The bank ends up with a record of who approved what, and that record holds up if anyone disputes it later.
Watch the full conversation, including the deepfake fraud cases weโre seeing and why Pat co-wrote an internet usersโ bill of rights back in 2007: hubs.ly/Q04FqwK10
202
United Airlines will block personal AI assistants from booking flights on its systems, citing customersโ sensitive data and purchases they never intended. ๐ซ
United can already tell when a booking comes from an AI agent. Before it lets one buy, an airline also needs to know which traveler the agent works for and what that traveler approved.
We built x401 so a site can verify the person behind an agent, and the limits that person set, before the agent buys anything.
Check out the link in the thread to read more about United's move, and how x401 can enable you to say YES to agentic commerce.
1
122
AI agents already shop for people. Soon they will open accounts and bind insurance policies, and the bank or insurer on the other end will have a valid API key and nothing else to go on. It won't know which person sent that agent or what they agreed to.
Fraud controls today assume a human is present to be checked. Device signals and document checks both depend on that person, and an agent acting for someone sidesteps them without doing anything that looks wrong.
@patk will be on the mainstage at @FintechNerdCon in San Diego on Nov 20 at 9:50am PT with fraud leaders from @SocureID and @Sardine, in a discussion moderated by @Baselayerhq's @jsawadd.
Then at 11:00am PT, Darren Louie and Daniel Buchner will run a hands-on x401 session in the Builder Zone. Bring a laptop.ย
๐ Link to the session details (and how you can connect with us during the show) in the thread.
3
4
184
Connect with the Proof team at Fintech Nerdcon: hubs.la/Q04F5b0x0
39
Proof retweeted
Replying to @ThePaypers @money2020
1/ Agentic commerce is moving from concept to execution, and the focus now is trust.
@circle and @proof are giving AI agents verified credentials to pay in USDC, and @evervaultโs Browser Tokens let agents complete purchases without seeing payment details.
A fraudulent wire shows up in the loss report within days. But a forged beneficiary change processes cleanly, updates the record, and pays the wrong person years later, often after the one person who could have disputed it has died.ย
Few recordkeepers can tell you how often it happens. The systems that count fraud are built around money leaving an account, and this change moves no money on the day it's filed. ๐ฌ
Pat Kinsel, our founder and CEO, will take on where retirement fraud is heading at SPARK Forum in Palm Beach on Monday, Nov 16 at 3:15pm ET. Heโll join Trish McGinity of Empower, Lisa Plaggemier of the National Cybersecurity Alliance and Jason Ruger of Lenovo, for a discussion moderated by Brendan Purcell of TIAA.ย
Plus, our team will be running demos in the exhibit hall. Come see us at booth 19 to tackle your most frustrating use cases!
Connect with the Proof team on site: hubs.la/Q04DMDtN0
156
When a purchase an agent made gets disputed, the first question is who approved it, and most records in the system today were built for a person clicking a buy button.
During Money20/20, our own @patk joined a panel hosted by @sytaylor, alongside leaders from Visa Intelligent Commerce, @circle, and @TDBank_US, to talk about what they need to see before agents get more room. The consensus: weโre missing a verified human authorizing agent action.
Thatโs exactly why we built x401, an open protocol that lets any website or API challenge an agent for verifiable proof of who it represents. It works for a wire transfer, an account opening, or an insurance purchase, anywhere an agent acts for a person.
Agents are already buying. So the time is now to build authorization before the disputes start rolling in.
๐ Watch the full panel discussion: hubs.la/Q04D687J0
90
On October 27, RBC blocked third-party AI agents from online banking. Within a day, TD, BMO, CIBC and Scotiabank moved to block or limit them too. Other banks are seeing agentic browsers clear income, employer and government ID checks, with a live selfie as the last check left. Thereโs no consensus yet on the right response, so weโre hosting an open conversation for banks about it.
On November 17, we'll be sharing what weโre hearing from banks, show what agentic workflows look like today, and take your questions on:
โข Tying agent payment tokens back to a verified identity
โข Account opening when the applicant may not be a person
โข What changes for corporate banking and treasury
Join the conversation: hubs.la/Q04DlYxr0
2
200
Merchants decide which agents get in, and they need a reason to say yes.
During our Money20/20 panel discussion, Visa Intelligent Commerceโs Ines Clark set the bar in two parts. The payment has to stay inside its guardrails, meaning the right category and the right spend amount, and it has to match the intent the person expressed. The agent also has to be verified as a trusted agent before the merchant lets it in.
x401 gives the merchant that check. A site asks the agent for proof of who it represents and a signed approval of the exact terms, and lets in the ones that can show both.
๐ Watch the full panel discussion: hubs.la/Q04D6cqs0
1
1
137
During our panel discussion at Money20/20, Haley Stelzer-Monahan of @TDBank_US shared what a bank needs before agents get more room: limits, guardrails, and auditability.
TD's corporate clients are already putting agents on invoice matching, reconciliation, forecasting, and chasing down missing remittance data. The agent prepares the payment, with a human still in the loop. As agents take on more of that work, every item on Haley's list becomes a record the bank has to be able to check.
A verified human and a signed mandate are two of those records, and x401 is how a site asks an agent for them.
๐ Watch the full panel discussion: hubs.la/Q04D635x0
2
99
A policy and a mandate are two different things, and @gaganmac drew the line with one example during our Money20/20 panel discussion.
A policy is a standing rule in a wallet: "spend up to $5,000, never more than $100 per transaction." A mandate is the instruction behind a single purchase: "I authorize you to spend up to $250 for getting this service."
Wallets already enforce policies well. In Gagan's words, "I think the explicit mandates are the missing piece."
With x401, the person signs the exact terms of the transaction, and if the amount or the recipient changes, the signature stops verifying. And paired with x402, the agent presents a verified credential and pays in USDC.
๐ Watch the full panel discussion: hubs.la/Q04CLJMp0
2
114
Days before #Money2020, we launched x401, an open protocol that lets any website or API ask an AI agent who it's acting for and what that person approved. It already works with Meta's Muse, ChatGPT, Grok and Claude.
During the event, we hosted an evening at CHICA in The Venetian and put that question to @TDBank_US, Visa Intelligent Commerce, and @Circle, in a panel hosted by @sytaylor.
A bank answers for disputed payments, a network decides which agents get in, and a wallet moves the money. All three need to know which verified person is behind the agent and exactly what that person approved.
Thank you to everyone who joined us! And if you're having FOMO, you can watch the full session on-demand: hubs.la/Q04CPtFk0
#AgenticCommerce #x401
2
167
Proof retweeted
The emerging standard x401 protocol fixes this --> x401.id
I'm shaking my head at "KYA". Banks and merchants don't want to know the agent, they want to know what the consumer permissioned and restrict the agents abilty to act within that authorization. Perhaps we should call it "register and restrict your agent". That is what Stripe, Walmart and Amex are all doing. Its all about agent control.. not about knowing them.
Agentic commerce is here. The proof of authority is missing.
Say an agent spends $250 on a service for you, and you dispute the charge a week later. Your bank asks who authorized it, the merchant has never dealt with that agent, and the network wants evidence, yet every one of them is working from a record built for a person clicking a buy button.
At #Money2020, we brought together leaders from
Visa Intelligent Commerce, @TDBank_US, @circle , and our own @patk for a panel discussion moderated by @sytaylor.
When asked to pick a side, block agents or allow them, the panel answered โWe are team authorized.โ โ
Agents are already clicking โbuy,โ so the disputes are coming whether the records are ready or not.
We'll have the full session available on-demand soon. In the meantime, check out the recap: hubs.la/Q04BVhJC0
A Worker gets a request from an AI agent. It canโt tell who sent it, or whether that person approved this. x401 now works with Cloudflare Workers: the Worker asks, the person approves in their Proof wallet, the Worker verifies. See it at #Money2020, Booth 5069: hubs.la/Q04BGT8R0
Circle ๐ค @proof
With Proof, weโre exploring how agents can prove who they act for and pay in USDC in one exchange.
Verified identity through x401. USDC payments through x402 and Circle Agent Wallets.
See the joint demo at @money2020, booth 5069.
globenewswire.com/news-releaโฆ










