Identi3

In this session, we show how an AI agent can be given a narrow, revocable slice of a person's authority, and how the counterparty on the other side of the transaction can verify it. Identifying an agent is not the hard part. Proving who authorized it, what they authorized, and whether the action falls inside those limits is.
The demo follows an office manager at a fictional enterprise who delegates part of her purchasing authority to an agent that reorders paper and toner. When the agent places the order, the vendor verifies the delegation chain: who granted the authority, what it covers, whether it is still valid, and whether this specific purchase sits inside it.
(0:00) Why agent identity alone is not enough
(2:10) Framing the problem: agents can transact, but cannot be trusted
(5:09) The three questions behind trusting an agent
(6:16) Delegation in four steps: issue, present, verify, revoke
(7:46) The scenario: Lisa, Atlas and Stockwell
(9:17) Demo: a cloud wallet embedded in the employee portal
(11:07) Purchasing authority as a credential with limits
(12:48) Delegating a narrow slice of purchasing authority
(13:57) The agent assembles and submits an order
(15:00) Verification from the vendor's side
(16:21) Audit log, and why agents are not privacy holders
(17:26) Changing the rules: schemas, rule sets and Cedar policy
(19:37) Capabilities: vendors, product categories and spend limits
(20:47) Delegation depth and sub-agent redelegation
(22:07) Using the same mechanism for human delegation
(23:26) Why an internal procurement use case
(25:31) Where the vendor's trust in the buyer comes from
(27:42) What the verification evidence actually contains
(30:13) Responsibility, accountability and the delegation chain
(32:07) Setup walkthrough: onboarding, issuance, delegation
(34:28) Transaction walkthrough and audit trail
(36:22) How much credential detail should a verifier ask for
(40:03) Technology used: REST API, Wallet SDK, MCP servers
(41:37) Agent wallets over MCP
(42:38) Landscape: bot blocking, OAuth extensions, TAP, KYA-OS, x401, Verifiable Intent
(46:43) Closed ecosystems versus industry-wide adoption
(48:31) Roadmap: standards alignment, auditing and reporting
(52:01) Payment settlement and AP2
(53:50) When delegation matters and when payment is enough
Website - https://www.dock.io/
LinkedIn - https://www.linkedin.com/company/docknetwork/

In this session, Andrew Shikiar, CEO at the FIDO Alliance, explains why passkeys succeeded where earlier password replacements failed, and how the FIDO Alliance is now applying the same playbook to digital credentials and AI agents.
(0:00) Introduction
(2:00) What the FIDO Alliance does, and where passkeys stand today
(5:04) Why passkeys succeeded where earlier password replacements did not
(9:03) Why the password problem unified the whole industry
(10:22) What a passkey proves, and what it does not
(11:07) Passkeys versus credentials, or passkeys and credentials
(12:32) Giving verifiable credentials a passkey moment
(14:44) eIDAS and the EUDI wallet as catalyst
(15:23) Inside the baseline wallet profile
(17:26) Will other regions copy the EUDI model?
(20:02) Derived credentials and the layer above the regulated core
(23:23) Where agentic commerce actually is today
(27:07) Why agentic commerce is a liability problem first
(29:12) Will B2B agentic commerce arrive first?
(33:47) What a merchant needs before it can trust an agent
(37:58) What actually changed in identity over the past decade
(39:40) AP2 and Verifiable Intent donated into FIDO
(42:16) What AP2 and Verifiable Intent each solve
(50:25) Audience question: how identity strategies are broadening
(52:16) Audience question: who gives an agent its identity
(53:18) Audience question: certification and smaller wallet providers
(56:46) Privacy in the wallet certification programWebsite - https://www.dock.io/LinkedIn - https://www.linkedin.com/company/docknetwork/

In this session, Harsh Mehta of Worldpay and Dock Labs unpack the identity and trust challenges behind AI agents and agentic payments: what is genuinely working today, what is still marketing, and what has to be solved before agents can transact autonomously.
(1:25) Meet the guest: Harsh Mehta on AI and agentic commerce at Worldpay
(5:47) Defining agentic commerce: reactive agents vs proactive agents
(8:08) The levels of autonomy and why the headlines run ahead of reality
(9:06) Human in the loop today: ACP, Google's protocol, and why browsers stop at checkout
(12:47) Are more agent transactions slipping through than we realize?
(13:18) Structured retail flows, embedded checkout, and why discovery is the long pole
(15:24) Why B2B and procurement may scale faster than consumer shopping
(16:50) The three hardest unsolved problems: data loss, identity, intent and liability
(19:20) How agentic commerce inverts twenty years of fraud detection
(20:54) False positives: when your fraud stack blocks your best new customers
(21:57) Dock Labs' model: tying agents to individuals through delegation and verifiable credentials
(25:44) Liability at agentic scale and who ends up holding the can
(27:56) Intent frameworks: Mastercard Verifiable Intent, Google AP2, and selective disclosure
(29:00) When bad product data is the real culprit
(32:27) Who owns the mandate verification layer: platforms, PSPs, or a neutral layer
(37:05) Commercial models: where value lands when an agent sits in the middle
(40:51) Closing takeaways: build the foundations now, do the unglamorous work
Website - https://www.dock.io/
LinkedIn - https://www.linkedin.com/company/docknetwork/

In this session, Agne Caunt and Richard Esplin from Dock Labs break down the EU Digital Identity Wallet (EUDI) at the level most explainers skip: how the ecosystem is actually structured, what matters versus what is noise, and the practical steps an organization can take today.
(00:00) Introduction
(06:18) Regulation: from eIDAS 2014 to the 2024 revision
(07:30) The BLT model: business, legal, and technical layers
(09:19) Where Europe innovated (legal and business) versus stayed conservative (technical)
(10:44) The three credential types: PID, QEAA, EAA
(12:18) Platform and format: when EUDI-compliant and "compliant enough" diverge
(14:39) Wallets: state wallets versus private wallets
(17:01) QEAAs in private wallets and what DICE revealed
(19:18) Google, Apple, and Samsung: where big tech wallets fit
(23:15) Where a technology provider fits across the ecosystem
(26:10) The standards, in one slide: OpenID4VC, HAIP, SD-JWT VC, mdoc
(29:47) What EUDI deliberately left out: JSON-LD, DIDs, and ZKPs
(33:34) An honest assessment of what is still moving
(35:14) Practical steps: the relying party path
(45:12) The issuer path: QEAA versus EAA
(47:48) The wallet provider path and white-labeling
(54:20) The European Business Wallet and the WE BUILD consortium
(57:19) Payment models: who pays whom
(1:01:24) Audience Q&A: country front-runners, multiple wallets, and international interoperability
Website - https://www.dock.io/
LinkedIn - https://www.linkedin.com/company/docknetwork/

What happens to liability when the entity making a purchase is not a human? AI agents are moving from research into commercial reality. They can research and execute purchases. But the legal, identity, and payment infrastructure that commerce runs on was designed around humans in the loop. This session explores what breaks when agents enter the picture and what needs to be built before agentic commerce can scale.
Guests:
- Przemek Praszczalek, Product Lead at Invela Network (formerly Mastercard, nine years in emerging payments and verifiable credentials)
- Ronald Kogens, Partner at MME, a Swiss law firm specializing in technology and financial markets law
(0:00) — Introduction and framing
(6:44) — How is agentic commerce different from traditional e-commerce?
(10:18) — Legal perspective: what changes when a non-human initiates a transaction?
(12:46) — Authenticating the agent vs. authenticating the person behind it
(16:26) — Bank liability when agents interact with fraudulent merchants
(19:07) — The closed vs. open ecosystem problem
(20:34) — Does the "I agree" button still work when an agent is clicking it?
(25:07) — Dispute flows, chargebacks, and cognitive overload
(28:31) — When the agent screws up: who is actually liable?
(31:33) — The EU AI Act and its implications for agent developers
(33:18) — New fraud vectors introduced by agentic commerce
(37:10) — Should agentic transactions be flagged differently in payment rails?
(40:08) — How will dispute resolution change in practice?
(43:37) — Could agents ever have legal personality?
(46:00) — Timeline: when does autonomous agentic commerce actually arrive?
(48:43) — B2B vs. B2C: which scales faster and why?
(54:16) — Audience Q&A: UK regulation and the T&Cs consent problem

Website - https://www.dock.io/
LinkedIn - https://www.linkedin.com/company/docknetwork/