Skip to main content
We have begun the founding chapter. Be among the first

TIP standard | Version 1.2

TIP for AI Agents

Verified agent identity, delegated authority and verifiable actions on the Trust Identity Protocol. Which agent acted, on whose authority, what it did, and whether a human approved it.

By Dinesh Mendhe for The AI Lab Intelligence Unobscured, Inc. Published November 13, 2026, 119 pages, CC BY 4.0. DOI 10.5281/zenodo.24211833

The four questions it answers

  1. Which agent acted?

    Every agent signs with its own key, bound to the verified TIP-ID of the person or organization that runs it.

  2. On whose authority?

    A signed mandate says what the agent may do, where and until when, and it only narrows authority its issuer already holds.

  3. What exactly did it do?

    The agent signs the exact bytes it sends, and the receiving site signs a receipt for every request, refusals included.

  4. Was a human there when it mattered?

    A high-risk action needs a fresh fingerprint, face or security key check on a device the agent cannot control.

What the standard defines

TIP Verified Agent and TIP Agent ID (ATID)

An agent's own signing key, bound to the verified TIP-ID of the person or organization that runs it, through a public Agent Record.

TIP Agent Mandate

A scoped, time-limited, revocable permission slip. It only narrows authority its issuer already holds at a site, and a helper agent can receive a narrower part of it, never a wider one.

TIP Approval

A fresh fingerprint, face or security key check on one exact high-risk action, shown with its amount, payee, record count or dose on a device the agent cannot control.

TIP Action Request

The agent's signature over the exact bytes, the target site and a one-time number, so a request cannot be altered or replayed.

TIP Receipt

The receiving site's signed decision for every request, refusals included, kept in a tamper-evident log.

TIP Seal

A witnessed summary of a site's receipt log, anchored on the TIP ledger in batches, so the ledger never stores individual actions.

TIP Stop All

A scoped, recoverable stop for every agent and mandate at once, which hands industrial processes back to manual control.

TIP Verifiable Action

An agent action that carries a TIP Receipt under a TIP Seal, which anyone can later check from public keys and the ledger.

Every object is signed with post-quantum ML-DSA-65 (FIPS 204), with an ML-DSA-87 profile for national security systems, and hashed with SHAKE-256. The paper sets out a threat model of 62 attacks with their remaining risks, safe-state operation for hospitals and industrial plants, profiles for HTTP Message Signatures, Web Bot Auth, OAuth 2.1, the Model Context Protocol, Agent2Agent, AP2, HL7 FHIR and SCITT, and a mapping to HIPAA, NIST, DORA, PSD2, the EU AI Act and other rules that names what each organization must still do itself.

How it reaches use

The parts that live off the TIP ledger form the TIP Agent Profile. These are mandates, approvals, signed action requests, receipts and site logs. The AI Lab publishes the profile and deploys it now as its own service.

The parts that change the TIP ledger follow the public TIP RFC process, because every node must run them. These are public agent records, ledger revocation, status signers, witnesses and anchoring.

Permanent record

The standard is permanently archived on Zenodo, operated by CERN.

Concept DOI, always the newest version
10.5281/zenodo.24097549

Cite it

APAMendhe, D. (2026). TIP for AI Agents: Verified Agent Identity, Delegated Authority and Verifiable Actions on the Trust Identity Protocol (Version 1.2). The AI Lab Intelligence Unobscured, Inc. 10.5281/zenodo.24211833

IEEED. Mendhe, “TIP for AI Agents: Verified Agent Identity, Delegated Authority and Verifiable Actions on the Trust Identity Protocol,” Version 1.2, The AI Lab Intelligence Unobscured, Inc., Wilmington, DE, 2026. doi 10.5281/zenodo.24211833

ChicagoMendhe, Dinesh. 2026. TIP for AI Agents: Verified Agent Identity, Delegated Authority and Verifiable Actions on the Trust Identity Protocol. Version 1.2. Wilmington, DE: The AI Lab Intelligence Unobscured, Inc. 10.5281/zenodo.24211833.

BibTeX

@techreport{mendhe2026tipagents,
  author       = {Mendhe, Dinesh},
  title        = {TIP for AI Agents: Verified Agent Identity, Delegated Authority and Verifiable Actions on the Trust Identity Protocol},
  institution  = {The AI Lab Intelligence Unobscured, Inc.},
  year         = {2026},
  month        = {11},
  version      = {1.2},
  doi          = {10.5281/zenodo.24211833},
  url          = {https://doi.org/10.5281/zenodo.24211833},
}

Version 1.2, published November 13, 2026. Version 1.1 added the rule for keeping evidence, the optional basis digest and the terms for hosted services, the same day. Version 1.0 was published November 9, 2026. Licensed under CC BY 4.0; implementations under TIPCL-1.0. Errata go to [email protected].

Questions

What is TIP for AI Agents?

TIP for AI Agents is an open standard, published by The AI Lab Intelligence Unobscured, Inc., that extends the Trust Identity Protocol from verified people and content to the actions of AI agents. It answers four questions about any agent action. Which agent acted, on whose authority, what exactly it did, and whether a human was there when it mattered. Version 1.2 was published on November 13, 2026 under DOI 10.5281/zenodo.24211833 and is licensed under CC BY 4.0.

Who wrote TIP for AI Agents?

Dinesh Mendhe, Founder and Chairman of The AI Lab Intelligence Unobscured, Inc., wrote it. His ORCID iD is 0000-0003-1158-3259. The paper builds on an internal research paper and reference implementation by Rama Chandra M of The AI Lab.

Can an organization use TIP for AI Agents today?

The parts that live off the TIP ledger form the TIP Agent Profile, which The AI Lab publishes and deploys now as its own service. These are mandates, approvals, signed action requests, receipts and site logs. A deployment still needs TIP clients that show and sign mandates, the reference code reading real TIP-IDs, and a way for each site to publish its receipt key, as Section 23.1 of the paper explains. Public agent records, ledger revocation, witnesses and anchoring change the TIP ledger, so they follow the TIP RFC process.

Does TIP for AI Agents make an agent safe or a company compliant?

No. TIP records which agent acted, on whose authority and what it did, limits what a compromised agent can reach, and leaves evidence nobody can edit away without detection. It does not stop prompt injection, does not decide whether an action was wise, and does not by itself make anyone compliant with HIPAA, the EU AI Act or any other law. Section 21 of the paper shows where TIP records can serve as evidence and what each organization must still do itself.

Where do prompts, request bodies and records live?

With the parties who own them. The TIP ledger holds agent records, revocations and Merkle roots, never request bodies, prompts or receipts. Each party keeps its own originals in its own systems and region (Section 12.8), and an optional salted basis digest lets an agent's operator commit to the instructions, model and tool results an agent acted on without revealing them (Section 10.6).

How does TIP for AI Agents relate to the TIP Protocol Whitepaper?

It is a companion to the TIP Protocol Whitepaper, Version 1.1 (DOI 10.5281/zenodo.22134504), which defines verified human identity and content provenance. TIP for AI Agents uses the same TIP-IDs, the same post-quantum ML-DSA-65 signatures and the same ledger, and adds the objects an agent needs to act for a person or an organization.

How do I cite TIP for AI Agents?

Mendhe, D. (2026). TIP for AI Agents: Verified Agent Identity, Delegated Authority and Verifiable Actions on the Trust Identity Protocol (Version 1.2). The AI Lab Intelligence Unobscured, Inc. https://doi.org/10.5281/zenodo.24211833