Declare · Enforce · Prove

AgenTrust
Agentic Controls

Thirty controls for AI agents, each stated as one requirement and cross-linked to the Common Requirement that already covers it in OpenCRE. Every control has a stable permalink so an external standard can cite it without the link rotting.

30 controls · 62 OpenCRE links · 53 Common Requirements
How to read this

The links point outward on purpose

Most of what an agent needs is not new. Identity, least privilege, logging, injection handling and rate limiting are decades old, and OpenCRE already organizes them across ASVS, CWE, ISO 27001, NIST 800-53, SAMM and the OWASP AI Exchange. So each control below names the Common Requirement it belongs to rather than restating it.

Where a control reads new, it is usually the same requirement applied to a different subject: a non-human principal, a grant derived from another grant, or a platform the relying party does not control. That distinction is what the OpenCRE links make visible.

This is a mapping, not a conformance claim. A link to a Common Requirement says a control belongs to that topic. It is not evidence that anything is implemented, tested, or deployed.

The control set

Thirty controls

Where a control is enforced at a specific point in the agent loop, the point is named on the control itself. Fourteen of the thirty have one. The rest do not have a single point and are left unmarked rather than forced into one.

E1 · Identity and authority

AT-01 Agent identity credential

agent_startup

The agent presents a credential bound to its own identity, distinct from the identity of the user who invoked it, before any authorization decision is made on its behalf.

Common Requirements: Authentication Authentication mechanism

Permalink: https://agentrust-io.com/go/agent-identity-credential/

AT-02 Agent key binding and custody

The agent's signing key is bound to its declared identity and held so that possession of the key cannot be separated from the identity it asserts.

Common Requirements: Secret storage Do not use static secrets

Permalink: https://agentrust-io.com/go/agent-key-binding/

AT-03 Capability attenuation across the delegation chain

pre_tool_call

A grant passed to a sub-agent is a subset of the grant it derives from. Delegation depth is bounded and no hop may widen scope.

Common Requirements: Minimize permissions Model action privilege minimization

Permalink: https://agentrust-io.com/go/capability-attenuation/

AT-04 Declared agent purpose and scope

The agent declares its purpose and the boundary of its task in a form a policy engine can read, not only a form a person can read.

Common Requirements: Documentation of all components' business or security function AI user transparency

Permalink: https://agentrust-io.com/go/agent-purpose-declaration/

AT-05 Declared capability manifest

agent_startup

The agent declares the tools and scopes it may use before it runs, and that declaration is the upper bound on what it can be granted.

Common Requirements: Model action privilege minimization Maintain/manage inventory of third party components

Permalink: https://agentrust-io.com/go/capability-manifest/

E2 · Behaviour and accountability

AT-06 Structured action logging

Every action an agent takes is recorded in a structured form carrying the actor, the action, the target, and the decision that permitted it.

Common Requirements: Log relevant

Permalink: https://agentrust-io.com/go/structured-action-logging/

AT-07 Signed, third-party-verifiable evidence record

The action record is signed so that a party who does not trust the operator of the log can still verify who acted and what was decided.

Common Requirements: Log integrity Audit & accountability

Permalink: https://agentrust-io.com/go/verifiable-evidence-record/

AT-08 Behavioural baseline for an agent

Normal behaviour for an agent is characterised before it is trusted with unattended work, so departure from it can be recognised.

Common Requirements: Detect and prevent unusual activity Monitoring

Permalink: https://agentrust-io.com/go/behavioural-baseline/

AT-09 Anomaly detection on agent behaviour

Departure from the established baseline is detected during the run rather than reconstructed after it.

Common Requirements: Monitor inference Anomalous AI input handling

Permalink: https://agentrust-io.com/go/agent-anomaly-detection/

AT-10 Policy verdict rationale

post_model_call

Each policy decision carries the rule that produced it and the inputs it read, so a verdict can be explained without re-running the agent.

Common Requirements: AI user transparency Automated AI oversight

Permalink: https://agentrust-io.com/go/policy-verdict-rationale/

E3 · Data and content

AT-11 Schema validation of agent input

input

Input reaching the agent is validated against a declared schema before it is used.

Common Requirements: Input validation

Permalink: https://agentrust-io.com/go/input-schema-validation/

AT-12 Prompt injection prevention

input post_model_call

Instructions arriving inside data are treated as data. Content fetched or returned during a run cannot alter the agent's instructions.

Common Requirements: Direct prompt injection Indirect prompt injection Prompt injection I/O handling Prompt input segregation

Permalink: https://agentrust-io.com/go/prompt-injection-prevention/

AT-13 Personal data protection in agent output

output

Personal data in the agent's context and output is identified and handled according to its classification rather than passed through.

Common Requirements: Personal data handling Sensitive AI output handling Data disclosure in model output

Permalink: https://agentrust-io.com/go/output-personal-data/

AT-14 Encoding and injection prevention

output

Model output is encoded for the interpreter that receives it, so output cannot become an injection in a downstream system.

Common Requirements: Output encoding and injection prevention Encode model output Model output contains conventional injection

Permalink: https://agentrust-io.com/go/output-encoding/

AT-15 Context provenance for agent working memory

What entered the agent's working memory, and from where, is recorded, so a poisoned context can be traced to its source.

Common Requirements: Augmentation data integrity controls Data supply chain management

Permalink: https://agentrust-io.com/go/context-provenance/

E4 · Scope and resources

AT-16 Resource allowlist

pre_tool_call

The data sources and sinks an agent may reach are declared in advance and enforced at the call boundary.

Common Requirements: Whitelist data sources and sinks

Permalink: https://agentrust-io.com/go/resource-allowlist/

AT-17 Tool authorization decision

pre_tool_call

Every tool call is an authorization decision made outside the model against the agent's granted scope, not a decision the model makes about itself.

Common Requirements: Model action privilege minimization Strong authorization checking Sanitization and sandboxing

Permalink: https://agentrust-io.com/go/tool-authorization/

AT-18 Rate limiting on agent actions

The rate at which an agent may act is bounded independently of the rate at which it may infer.

Common Requirements: Rate limiting against AI input attacks Limit inference resources Denial Of Service protection

Permalink: https://agentrust-io.com/go/agent-rate-limiting/

AT-19 Transaction and spend limits

pre_tool_call

Transaction value and cumulative spend are bounded per run and per agent, and enforced before the call rather than reconciled after it.

Common Requirements: Impact limitation of unwanted model behaviour Enforce additional authorization and segregation of duties

Permalink: https://agentrust-io.com/go/transaction-limits/

AT-20 Blast radius containment for agent execution

An agent executes inside a boundary that limits what a compromised or mistaken run can reach.

Common Requirements: Sandbox, containerize and/or isolate applications at the network level Impact limitation of unwanted model behaviour

Permalink: https://agentrust-io.com/go/blast-radius-containment/

E5 · Response and recovery

AT-21 Circuit breaker on agent loops

post_tool_call

Repeating or non-terminating agent loops are broken automatically, and a broken loop does not cascade into the agents depending on it.

Common Requirements: Impact limitation of unwanted model behaviour Fail securely

Permalink: https://agentrust-io.com/go/agent-circuit-breaker/

AT-22 Terminate a running agent

agent_shutdown

A running agent can be stopped by an operator, and the stop takes effect on work already in flight.

Common Requirements: Human AI oversight Incident response

Permalink: https://agentrust-io.com/go/agent-termination/

AT-23 Agent session revocation

An agent's session and the authority attached to it can be revoked without waiting for expiry.

Common Requirements: Minimize session life Enable option to log out from all active session

Permalink: https://agentrust-io.com/go/agent-session-revocation/

AT-24 State rollback after an agent action

State an agent changed can be identified and reversed, distinctly from restoring a backup of the whole system.

Common Requirements: Backup Change management

Permalink: https://agentrust-io.com/go/agent-state-rollback/

AT-25 Graceful degradation on policy engine failure

When the policy engine is unavailable the agent degrades to a defined and safe behaviour rather than to an unenforced one.

Common Requirements: Fail securely

Permalink: https://agentrust-io.com/go/graceful-degradation/

A · Attested primitives

AT-26 Runtime attestation evidence

agent_startup

Evidence of what is actually executing is produced by the platform at startup and appraised by a relying party that does not control that platform.

Common Requirements: Allow only trusted sources both build time and runtime; therefore perform integrity checks on all resources and code Runtime model integrity controls

Permalink: https://agentrust-io.com/go/runtime-attestation-evidence/

AT-27 Transparency-log anchoring of evidence

Evidence records are anchored in an append-only log so a record cannot be rewritten after the fact without detection.

Common Requirements: Log integrity

Permalink: https://agentrust-io.com/go/evidence-transparency-anchoring/

AT-28 Attested agent-to-agent channel

An agent-to-agent channel establishes what the peer is, not only that the channel is encrypted, before scope is passed across it.

Common Requirements: Communication authentication Mutually authenticate application components

Permalink: https://agentrust-io.com/go/attested-a2a-channel/

AT-29 Continuous usage control after grant

Authorization is re-evaluated for the life of a grant, and obligations attached to a grant survive the moment it was issued.

Common Requirements: Strong authorization checking Minimize session life

Permalink: https://agentrust-io.com/go/continuous-usage-control/

AT-30 Model weight custody against the hosting operator

Model weights released into infrastructure the model owner does not control remain under the owner's custody, including against the infrastructure operator.

Common Requirements: Runtime model confidentiality controls Model hosting supply chain mannagement AI model supply chain management

Permalink: https://agentrust-io.com/go/model-weight-custody/

Limits

What this page does not do

This page is generated from the OpenCRE mapping sheet. Edit the mapping, not this file.