GuardianChain
AI governance

AI takes action.
Who authorized it?

Knowing which AI system performed an action is only part of the picture. Organizations require evidence of who authorized it, what was permitted, and what happened. GuardianChain brings these elements together in a record that can be inspected and verified.

See how it works Build with us

Design-partner stage

From permission to review
01Permission

A person sets the limits.

Who
The person or team giving permission
What
The task AI is allowed to carry out
Limits
What it can do, and for how long
02
AI action

Record what actually happened.

03
Review

Did the action stay within those limits?

Illustration of the proposed approach. People make the final decision.
01

Who authorized?

We record who authorized the AI to act.

02

What was allowed?

We capture the scope, limits, and duration of that permission.

03

What was done by AI?

We link the AI’s actions to their recorded outcomes.

The challenge of AI governance

Five questions every organization needs to answer.

  1. Authority

    Who authorized the AI to act, what was permitted, and for how long? A prompt alone is not a clear record of permission.

    Record who granted permission, what the AI is allowed to do, and how long that permission lasts.

  2. Durable proof of changes

    Logs held by the tool that did the work are not a record the recipient can lean on.

    Link each AI action to its authorization and supporting evidence, creating a record the recipient can inspect.

  3. Continuity across tools

    Work moves between assistants, editors and services. Context and limits are lost at each hop.

    Carry authorization records across connected tools, keeping the original scope and limits visible. Coverage depends on integration.

  4. Independent verification

    The party receiving the work needs to inspect it without trusting the producer's summary.

    Provide a portable record recipients can check independently. Verification supports their decision; it does not replace acceptance.

  5. Audit and rollback

    Security and incident teams need evidence they can reconstruct from, and a path back.

    Record when permissions expire or are revoked, supporting audits and incident review. Guardian Chain does not perform rollback.

The chain we connectArchitecture overview

Four links. One evidence chain.

GuardianChain connects who authorized an action, its permitted scope, the consequential decision, and independently verifiable evidence of the outcome.

  1. Who allowed this?Human authority

    A named, accountable person or role grants the action. A prompt or API key is not a grant.

    Record contents: the accountable grantor and the grant itself.

    Without it: an agent acts on a prompt or credential, and no one can show who allowed it.

  2. Within what limits?Bounded scope

    What may be done, on which resources, under which conditions, and until when. Refusals and restraint are recorded too.

    Record contents: the permitted actions, resources, conditions and expiry.

    Without it: authority silently renews or widens across tools and sessions.

  3. What was decided and done?Machine action

    The consequential decision made by an agent or automated system, linked to the authority and scope it relied on.

    Record contents: the consequential decision, linked back to its grant and scope.

    Without it: logs show what ran, but not whether it was allowed.

  4. Can someone else check it?Outcome evidence

    A portable record a recipient can inspect independently, without trusting the producer's own summary.

    Record contents: evidence references a recipient can inspect separately.

    Without it: recipients accept work on the producer's word.

Why existing tools are not enough

Existing tools don’t provide the full picture.

Identity, orchestration, tool connectivity, policy, and logging all matter, and we build on them. However, none of these systems alone establishes a complete chain from human authorization to a
verifiable outcome.

Typical coverage by category, in isolation. Individual products vary; this is our framing, not a benchmark.
Criteria Identity providers Agent orchestration Tool connectivity Policy engines Logs and observability GuardianChain architecture
Primary job Authenticate people, services and agents Plan and run multi-step agent work Connect agents to tools and data Evaluate rules at decision time Record events inside each system Connect everything, alongside the systems above
Who authorized Partial coverage Requires other tools Requires other tools Requires other tools Requires other tools Architectural focus
Permitted scope Requires other tools Partial coverage Partial coverage Yes Requires other tools Architectural focus
Decision linked Requires other tools Yes Yes Partial coverage Partial coverage Architectural focus
Independently checkable Requires other tools Requires other tools Requires other tools Requires other tools Partial coverage Architectural focus
Product inventory

Three components. One connected system.

Three architectural roles, with workflow fit and availability qualified through design-partner discussions.

Canonical authority and evidence substrate

GC-1

The record of who authorized what, within which scope, and what evidence exists for the outcome. The other parts depend on it.

How GC-1 works
Human and machine interface for governed sessions

GuardianBridge

Where a person grants, bounds, watches and ends a machine session, so authority is explicit at the moment it matters.

How GuardianBridge works
Portable, independently verifiable evidence

FactReceipt

A statement a recipient can check without trusting the producer: what was claimed, by whom, under what authority, with what support.

How FactReceipt works
Complements, never replaces

Built to work alongside your existing systems.

Built with established tools.

  • Website
    tools
  • Email
    connector
  • Listed
    technologies
Identity providerWe rely on your existing identity systems to know who someone is.
Secrets managerCredentials stay where you keep them today.
Agent frameworkYour agents and orchestration remain your choice.
Universal gatewayWe do not need to sit in every request path.
Compliance dashboardWe do not certify compliance or legal sufficiency.
Our technology ecosystem

Explore the tools used for this website, the email connector attached to our workspace, and the technologies listed in our wider directory. Directory entries do not indicate live integrations.

Explore all 33 tools and technologies

Website tools (4)

Tools used to build and operate this informational site.

Email connector (1)

Attached to the workspace. Email sending is not implemented.

Listed technologies (28)

Referenced in the GuardianChain technology directory. Not a live integration on this site.

Industry applications

AI governance across industries.

Software development workspace with code on computer monitors
Technical and AI deliveryDelivery acceptance when agents did part of the work.
Microscopes arranged along a laboratory bench
Regulated research handoffsProvenance and authority travel with materials. Not a compliance claim.
Aerial view of shipping containers and freight lanes at a port
Freight evidence reconciliationPreserved evidence when parties disagree about a shipment.
Vehicle assembly line with automated manufacturing equipment
Industrial operationsWhere automated decisions have physical consequences.
Research and regulation

Why AI actions need verifiable evidence.

Controlled research, not a forecast of routine failure. Read the sources and their limits

  1. Controlled scheming evaluations

    OpenAI's o1 System Card reported basic in-context scheming in deliberately constructed Apollo Research evaluations, not normal deployment rates.

  2. High-stakes simulations

    Anthropic's summer 2026 research documented covert sabotage, fraud assistance and motivated mislabeling in simulations, not real incidents.

  3. Real systems, in evaluations

    Anthropic reported four incidents of unauthorized access to real third-party systems during misconnected cyber evaluations. Not customer deployments.

  4. Lifecycle obligations

    EU AI Act high-risk requirements include logging, documentation and human oversight on a 2027 to 2028 timetable. No certification or automatic compliance is claimed here.