Appendix B: AI Security Controls Inventory

September 18, 2026 ยท View on GitHub

Objective

This appendix is a consolidated, developer-facing inventory of the security controls mandated across the AISVS requirements. Controls are grouped by control family so an implementer can find all related defenses in one place, regardless of which chapter defines them, and each control links back to the AISVS requirement IDs that mandate it.

This inventory is non-normative. It reorganizes existing requirements for ease of implementation and does not add, remove, or change any requirement. The requirement chapters (C1 through C12) remain the source of truth. Requirement IDs are written in canonical C{chapter}.{section}.{requirement} form (for example, C5.1.1). Every numbered requirement in the standard appears in exactly one control family below, so the inventory can be checked for completeness against the chapters.


AD.1 Authentication & Identity

Verify the identity of users, agents, services, edge devices, and MCP clients/servers before granting access.

Control / TechniqueRequirement IDs
Step-up authentication for high-risk AI operations (model deployment, weight export, training-data access, production configuration changes)C5.1.1
Short-lived, minimal-scoped, cryptographically signed tokens for federated or multi-system agent authenticationC5.1.2
Strong authentication of edge AI devices to central infrastructureC4.3.1
Unique cryptographic identity per agent instance, authenticating as a first-class principal to downstream systemsC9.4.1
Scheduled rotation of agent identity credentialsC9.4.3
MCP per-request access-token validation (not transport security alone)C10.2.1
MCP access-token claim validation (issuer, audience, expiration, scope) per OAuth 2.1C10.2.2
MCP resource servers do not store or persist access tokens or user credentialsC10.2.3
Removal of all MCP session artifacts on session terminationC10.2.6
MCP servers accept only tokens explicitly issued for themC10.2.7
Sender-constrained MCP access tokens (mTLS or DPoP)C10.3.5

Common pitfalls: reusing end-user credentials for agent-to-agent calls; not rotating agent credentials on suspected compromise; treating transport security as a substitute for per-request token validation.


AD.2 Authorization & Access Control

Enforce access decisions across users, agents, tools, and resources using policy that the model cannot override.

Control / TechniqueRequirement IDs
Access controls on every AI resource (datasets, endpoints, vector collections, embedding indices, compute) with explicit allow-lists and default-denyC5.2.1
End-user authorization context enforced at each retrieval and assembly stage, not the service account aloneC5.2.2
Post-inference filtering so responses exclude data the requester is not entitled to receiveC5.2.4
Policy decision point isolated from the agent execution environmentC5.2.5
Just-in-time privileged access to model weights, training pipelines, and production configuration with automatic expiryC5.2.6
Fine-grained, runtime-enforced authorization of agent actions (which tools, which parameter values)C9.5.1
Integrity-protected, scope-limited delegation token propagated to every downstream callC9.5.2
Access-control decisions enforced by application logic or a policy engine, never by the modelC9.5.3
Inter-agent task delegation restricted by an explicit authorization policyC9.5.5
Re-evaluation of backend authorization on every privileged action in long-running sessionsC9.5.6
Delegated authority never exceeding the delegating agent's current authorityC9.5.7
Scope-filtered MCP tool discovery (tools/list returns only authorized tools)C10.2.4
Per-invocation MCP access control validating both the tool and the supplied argument valuesC10.2.5
MCP proxy servers sharing an upstream OAuth client identity do not let one MCP client inherit another client's authorizationC10.4.9

Common pitfalls: relying on the service account's permissions instead of the caller's; letting model-generated output drive authorization; not re-checking authorization when context changes mid-session.


AD.3 Data Classification & Tenant Isolation

Keep data within its authorization and tenancy boundaries as it flows through AI-specific transformations and shared infrastructure.

Control / TechniqueRequirement IDs
Sensitive data served through retrieval pipelines rather than persisted into model weightsC5.2.3
Classification labels propagated to downstream resources (embeddings, prompt caches, model outputs)C5.2.7
Cross-tenant isolation in shared model serving (fine-tuning, inference, embedding operations)C5.3.1
Cross-tenant isolation across shared compute (hardware partitioning, confidential computing, or dedicated allocation)C5.3.2

Common pitfalls: dropping classification labels when data is embedded or cached; assuming logical multi-tenancy is sufficient against side channels in shared inference caches.


AD.4 Encryption & Data Protection

Protect data and secrets at rest, in transit, and in the model's observable context.

Control / TechniqueRequirement IDs
Integrity protection of training data while stored and transferredC1.1.3
Redaction, anonymization, or encryption of sensitive information in labels before use in any labeling artifactC1.2.3
Encryption of locally stored model weights and sensitive parameters using hardware-backed key stores or secure enclavesC4.3.4
Encryption at rest of models packaged in mobile, IoT, or embedded apps, decrypted only inside a trusted runtime or secure enclaveC4.3.5
Secrets and credentials kept out of the model's observable context (context window, system prompts, tool-call parameters)C9.5.4

Common pitfalls: encrypting the database but not model checkpoints or embeddings; leaving model weights extractable from an app package; exposing API keys inside tool-call parameters.


AD.5 Integrity, Signing & Provenance

Verify authenticity and detect tampering of models, artifacts, messages, tool definitions, and generated media.

Control / TechniqueRequirement IDs
Integrity monitoring of training data against unauthorized modification or corruptionC1.1.4
Cryptographic integrity for labeling artifactsC1.2.2
Cryptographic signing of all model artifacts (weights, configs, tokenizers, base models, fine-tunes, adapters, safety/policy models)C3.1.2
Signature verification at deployment admission and on loadC3.1.3
Signed edge/mobile model packages with on-device signature or checksum validation before loadC4.3.2
Cryptographic binding of agent-initiated actions to each step of the execution chain for non-repudiationC9.4.2
Integrity protection of agent state persisted between invocationsC9.4.4
Signed MCP tool responses with a unique nonce and timestamp for replay defenseC10.4.6
Tool-definition snapshotting with re-approval required on any change before invocationC10.4.8
Watermarking of AI-generated media to prove it was AI-generatedC7.4.4

Common pitfalls: using mutable tags instead of immutable digests; not re-verifying tool definitions between MCP invocations; missing replay protection on tool responses.


AD.6 Input Validation & Sanitization

Validate, normalize, and constrain all inputs (including tool, MCP, and retrieved content) before they reach the model or downstream systems.

Control / TechniqueRequirement IDs
Input normalization applied before tokenization or embeddingC2.1.1
Encoding and representation-smuggling detection and mitigation (canonicalization, strict schema validation, policy-based rejection, or explicit marking)C2.1.2
Untrusted-input screening by a prompt-injection detection ruleset or classifier, with blockingC2.1.3
Input length controls that reject (not truncate) content exceeding the context windowC2.1.4
Allow-list character-set restriction on all inputsC2.1.5
Instruction hierarchy enforcement (system and developer messages override user and untrusted input)C2.1.6
Reserved special tokens encoded as literal characters and not injectable into contextC2.1.7
Many-shot jailbreaking pattern detectionC2.1.8
Adversarial-perturbation, steganography, and hidden-content checks on non-text inputs (image, video, audio)C2.2.3
Cross-modal coordinated attack detectionC2.2.4
Schema validation of tool outputsC9.3.2
Verification of external resources named in model output against an approved allow-list or registry before install or invocationC9.3.7
MCP response schema validation before injection into model contextC10.4.1
Indirect-prompt-injection screening of MCP responses before injection into model contextC10.4.2
Rejection of unrecognized or oversized MCP function-call parametersC10.4.3
Strict MCP schema validationC10.4.4
Maximum MCP payload size limitsC10.4.5
Anomaly detection on external or untrusted inputs before inferenceC11.4.1
Gating actions on inputs flagged as anomalousC11.4.2

Common pitfalls: validating only the text modality while ignoring image/audio channels; relying on regex alone without semantic detection; not validating tool and MCP outputs before they re-enter agent context.


AD.7 Inbound Content & Policy Screening

Screen prompts and training content against policy before they reach the model or the training pipeline.

Control / TechniqueRequirement IDs
Inbound content classification (violence, self-harm, hate, sexual) against configurable thresholds, with rejection or sanitization before model contextC2.2.1
Evaluation of content classification for unsupported languagesC2.2.2
Detection and removal of disallowed content before trainingC1.3.4

Common pitfalls: deploying classifiers tuned only for one language; screening prompts but not the training corpus.


AD.8 Output Handling & Safety

Constrain, filter, and validate model outputs before they reach users or downstream systems.

Control / TechniqueRequirement IDs
Schema validation of model outputs with rejection on mismatchC7.1.1
Length limits and termination controls on generated outputC7.1.2
No action taken when the AI response used to decide it is a refusal, a reported failure, or a response the application cannot readC7.1.3
Confidence or uncertainty estimation for generated answersC7.2.1
Automatic blocking or fallback when confidence drops below a defined thresholdC7.2.2
Additional verification step for responses classified as high-risk by policyC7.2.3
Automated classifiers that scan responses and block defined harmful-content categoriesC7.3.1
Detection and blocking of responses that disclose system prompt content or backend dataC7.3.2
Prevention of model-generated output triggering outbound requestsC7.3.3
Detection of hidden, encoded, or misleading output (homoglyphs, formatting, metadata, structured fields)C7.3.4

Common pitfalls: enforcing stop sequences in batch mode but not on streaming output; leaking the system prompt through paraphrase; treating a confidence score as available when the provider does not expose one.


AD.9 Rate Limiting, Budgets & Resource Control

Bound consumption to prevent abuse, runaway execution, denial of service, and model extraction.

Control / TechniqueRequirement IDs
Per-tool quotas and timeouts (CPU, memory, disk, egress, execution time)C9.1.1
Per-execution budgets (maximum recursion depth, token use, monetary spend) enforced by the runtimeC9.1.2
Per-principal and global inference rate limits sized to the extraction threat model, not a generic API throttleC11.2.2

Common pitfalls: rate-limiting per endpoint but not per agent session; ignoring tool fan-out when sizing budgets; treating extraction defense as ordinary throttling.


AD.10 Sandboxing & Workload Isolation

Isolate models, tools, agents, and hardware workloads to contain failures and prevent lateral movement.

Control / TechniqueRequirement IDs
Execution of AI models in isolated sandboxesC4.1.1
Allow-list of serialization formats that do not permit code execution during deserializationC4.1.2
Workload attestation before model loadingC4.1.3
Confidential inference protecting model weights at runtime through isolated executionC4.1.4
Trusted execution environment with hardware-enforced isolation, memory encryption, and integrity protectionC4.2.2
GPU integrity validation via hardware attestation before each workloadC4.2.3
GPU memory partitioning with sanitization between jobsC4.2.4
Version-pinned, signed, boot-attested accelerator firmwareC4.2.1
Process, memory, and file-access isolation in edge inference runtimesC4.3.3
Least-privilege sandbox or isolation for each tool or pluginC9.3.1
Tool manifests declaring required privileges, resource limits, and output-validation requirementsC9.3.3
Runtime enforcement of declared tool-manifest privileges and limitsC9.3.4
Isolation of untrusted-data processing from tool-calling capabilityC9.3.5
Architectural separation of untrusted tool-output processing from agent operationsC9.3.6
Least-privilege sandbox for locally launched MCP servers (restricted file system, network, system access)C10.1.3
AI-specific runtime components not shared across environment boundaries (development, staging, production)C3.4.1
Training and fine-tuning environments isolated from productionC3.4.2

Common pitfalls: sharing infrastructure between dev and prod; granting tool sandboxes more capability than needed; allowing untrusted data processing to reach tool-calling paths.


AD.11 Network & Egress Control

Control network boundaries, transport security, and traffic flow for AI workloads and MCP integrations.

Control / TechniqueRequirement IDs
Authenticated, encrypted streamable HTTP for remote MCP transportC10.3.1
stdio MCP transport restricted to controlled local environmentsC10.3.2
Independent Origin and Host header validation on HTTP-based transports (DNS rebinding defense)C10.3.3
MCP client minimum protocol-version enforcement (downgrade defense)C10.3.4
Accelerator interconnects restricted to approved topologies and authenticated endpointsC4.2.5

Common pitfalls: exposing stdio or SSE transports beyond the local host; skipping Origin/Host validation and enabling DNS rebinding; accepting downgraded protocol versions.


AD.12 Supply Chain & Artifact Integrity

Verify origin and authenticity of models, datasets, frameworks, and MCP components, and maintain an AI bill of materials.

Control / TechniqueRequirement IDs
Model registry inventory of all deployed model artifacts and their originC3.1.1
Malicious-code scanning of models before importC6.1.1
Approved-source-only download of model weights, datasets, and fine-tuning adaptersC6.1.2
Integrity verification of every third-party model artifactC6.1.3
Behavioral acceptance test suite passed before promotion beyond developmentC6.1.4
Version-controlled, machine-readable AI BOM per model artifact (datasets, weights, licenses, data-origin statements)C6.2.1
Cryptographic signing of AI BOMs before deploymentC6.2.2
Build-failing AI BOM completeness checks when component metadata is missingC6.2.3
MCP components obtained only from trusted sources and cryptographically verifiedC10.1.1
Allow-listed MCP servers onlyC10.1.2

Common pitfalls: treating AI BOMs as static documents rather than signed, version-controlled artifacts; not scanning pretrained weights for backdoors; pulling models from unapproved registries.


AD.13 Model Lifecycle, Deployment & Rollback

Manage model validation, deployment, rollback, and fine-tuning pipeline integrity.

Control / TechniqueRequirement IDs
Pre-deployment automated input-validation, safety-evaluation, and output-sanitization testingC3.2.1
Re-evaluation of models subjected to post-training quantization against the same safety and alignment test suite before deploymentC3.2.2
Security re-evaluation triggered by provider model, version, or routing changesC3.2.3
Rollout mechanisms with automated rollback triggersC3.3.1
Complete model-state restoration on rollbackC3.3.2
Isolated runtime state for model versions running in parallelC3.3.3
Versioned, integrity-verified RLHF reward models before a training runC3.5.1
Detection of reward hacking or reward-model over-optimization in RLHF stagesC3.5.2
Stage-by-stage integrity verification in multi-stage fine-tuning pipelinesC3.5.3
Fine-tuning checkpoints registered as distinct artifactsC3.5.4

Common pitfalls: not testing rollback before it is needed; leaving retired model artifacts in serving caches; treating reward models as static infrastructure rather than versioned, validated artifacts.


AD.14 Training Data Integrity & Governance

Source, vet, and document training data so tampering, poisoning, and corruption can be detected and traced.

Control / TechniqueRequirement IDs
Data minimization to only the features, attributes, and fields required for the stated purposeC1.1.1
Up-to-date inventory of every training-data source (origin, responsible party, license, collection method, use constraints, processing history)C1.1.2
Dataset watermarking for usage attribution and detection of unauthorized useC1.1.5
Labeling-platform access controls restricting who can create, modify, or approve annotationsC1.2.1
Poisoning detection in training and fine-tuning pipelinesC1.3.1
Confidence thresholds and consistency checks on automatically generated labelsC1.3.2
Bias evaluation for models used in security-relevant decisionsC1.3.3
Defenses against clean-label poisoning attacksC1.3.5
Dataset lineage recording (transformations, augmentations, merges)C12.5.1
Logging of all labeling activitiesC12.5.2
Write-time tagging of every ingested document (source, writer identity, timestamp)C12.5.4

Common pitfalls: not scanning fine-tuning datasets for poisoning; collecting more attributes than the purpose requires; losing dataset lineage across transformations and merges.


AD.15 Memory, Embeddings & RAG Security

Harden vector stores, memory pipelines, and retrieval-augmented generation against leakage, poisoning, and fabricated provenance.

Control / TechniqueRequirement IDs
Per-tenant uniqueness of vector identifiers and namespaces, preventing cross-tenant collisionsC8.1.1
Immutability of document metadata tags after the initial writeC8.1.2
Scope constraints enforced on retrieval operationsC8.1.3
Detection and masking, tokenization, or dropping of sensitive fields before embeddingC8.2.1
Detection, rejection, or quarantine of retrieval-manipulation content before vectorizationC8.2.4
Flagging and quarantine of outlier vectors before they enter production indicesC8.2.2
Source validation before agent or tool outputs are written to trusted memoryC8.2.3
Contradiction checks on new memory writes, with conflicts triggering alertsC8.2.5
Exclusion of expired vectors from retrieval resultsC8.3.1
Memory reset capabilityC8.3.2
Retention of quarantined content while excluding it from all retrieval resultsC8.3.3
Attribution of RAG responses to their source documentsC7.4.1
RAG attributions derived from retrieval metadata, not generated by the modelC7.4.2
Traceability of RAG claims to the retrieved chunkC7.4.3

Common pitfalls: auto-writing tool output into trusted memory without validation; serving expired or quarantined vectors; letting the model fabricate citations instead of deriving them from retrieval metadata.


AD.16 Adversarial Robustness & Privacy Defense

Test for and defend against evasion, membership inference, model inversion, extraction, and poisoning of the improvement loop.

Control / TechniqueRequirement IDs
Alignment and safety training or fine-tuning to suppress disallowed content categoriesC11.1.1
Version-controlled alignment test suite run on every model update or releaseC11.1.2
Evaluation against known adversarial attack techniques relevant to the modalityC11.1.3
Hardening of models against adversarial inputsC11.1.4
Automated evaluator that measures harmful-content rate and flags regressions beyond a thresholdC11.1.5
Suppression of directly returned model-inferred sensitive attributesC11.2.1
Output calibration to reduce overconfident predictions exploitable by inference attacksC11.2.3
Differentially-private optimization for training on sensitive datasetsC11.2.4
Membership-inference attack simulation demonstrating accuracy no better than random guessingC11.2.5
Raw model outputs not exposed beyond the backend, with externally visible responses calibrated to extraction riskC11.3.2
Model watermarking or fingerprinting so unauthorized copies can be identifiedC11.3.3
Server-side constraints on caller-supplied sampling and decoding parameters, with out-of-policy values rejectedC11.3.5
Poisoning detection and human review gates protecting the safety-violation feedback pipelineC11.4.3

Common pitfalls: testing only known jailbreak patterns without adaptive attacks; not re-running the alignment suite after model updates; exposing raw confidence vectors that accelerate extraction.


AD.17 Logging & Audit

Capture security-relevant events with sufficient context and integrity for forensic reconstruction and accountability.

Control / TechniqueRequirement IDs
AI interaction logging with session context and AI-specific telemetryC12.1.1
Logging of safety filtering and policy decisions in enough detail to audit content moderationC12.1.2
Structured, interoperable log schema for inference events (model identifier, token usage, provider, operation type)C12.1.3
Logging of RAG pipeline retrieval events (query, documents retrieved, knowledge source)C12.1.4
Audit logs capturing the approval chain for security-critical proactive actions (approver identity, timestamp, parameters, outcome)C12.4.2
Logging of kill-switch activations and override commandsC12.4.3
Immutable audit records for all model changesC12.5.3

Common pitfalls: logging prompts without redaction; using mutable log storage without integrity protection; logging agent actions and approvals but not human-initiated overrides such as kill-switch activations.


AD.18 Monitoring, Detection & Incident Response

Detect AI-specific abuse, drift, and anomalies, and respond to incidents.

Control / TechniqueRequirement IDs
Automated tool containment triggered by policy violationsC9.3.8
Extraction-attempt detector fed by query-pattern analysisC11.3.1
Response measures triggered on detection of suspected model extractionC11.3.4
Signature-based detection and alerting on jailbreak patterns, prompt injection, and adversarial inputsC12.2.1
Behavioral anomaly detection (unusual conversation patterns, excessive retries, systematic probing)C12.2.2
Custom detection rules for AI-specific threat patterns (coordinated jailbreak attempts, prompt injection, system prompt extraction)C12.2.3
Extraction-alert events including offending query metadataC12.2.4
Granular token-usage attribution (per user, session, feature endpoint, team or workspace)C12.2.5
Monitoring of LLM API traffic for covert-channel and command-and-control indicatorsC12.2.6
Data drift detection using methods matched to the input type (KS test or PSI for tabular, embedding-distance for text/image)C12.3.1
Hallucination detection monitoring of model outputsC12.3.2
Hallucination rates tracked as continuous time-series metricsC12.3.3
Distinction of unexplained behavioral shifts from gradual operational driftC12.3.4
Security evaluation and threat-landscape assessment for autonomous action triggersC12.4.1

Common pitfalls: not correlating AI-specific events with broader SIEM alerts; treating drift as a scheduled check rather than continuous monitoring; lacking AI-specific forensic tooling during an incident.


AD.19 Human Oversight & Shutdown Control

Require human approval for high-impact actions and provide reliable, exercised shutdown and graceful-degradation paths under human control.

Control / TechniqueRequirement IDs
Swarm-level kill-switch that halts all active agent instancesC9.1.3
Runtime blocking of privileged or irreversible actions until explicit human approval is received and verifiedC9.2.1
Approval requests displaying canonicalized, complete action parameters (diffs, commands, recipients, amounts, resources, scopes) without truncationC9.2.2
Trusted reversibility classification for each high-impact action (read-only, reversible, externally reversible, irreversible)C9.2.3
Runtime enforcement of reversibility classifications (block, require approval, or restrict)C9.2.4
Restriction and bounding of any self-modification capability (prompt rewriting, tool-list changes, parameter updates)C9.2.5
AI-augmented review of planned high-risk actions, adding to (not replacing) the deterministic policy gateC9.2.6
Protection of the AI-augmented review mechanism against prompt-injection bypassC9.2.7
Approvals cryptographically bound to parameters, requester identity, execution context, and a single-use nonceC9.2.8
Isolation of approval-issuing key material or credentials from the agent runtimeC9.2.9
Approval gates enforcing the highest-impact reversibility classification across action chains established outside the gated agent's controlC9.2.10
Blocking execution beyond an approved action chain until new approval enforces the highest-impact reversibility classification across the cumulative chainC9.2.11
Manual kill-switch to immediately halt model inference and outputsC9.6.1
Fail-closed blocking of a pending action when a human-approval gate is not satisfied within the defined timeC9.6.2
Kill-switch commands delivered through an out-of-band channel isolated from the agent runtimeC9.6.3
Explicit consent dialogue and cancellation option on installation of a local MCP serverC10.4.7
MCP client consent and authorization bound to the approved server connection endpoint, with user re-approval before further interaction after an endpoint changeC10.4.10

Common pitfalls: documenting a high-risk action policy never wired to a runtime gate; binding approval to parameters without binding to identity or context; defaulting to fail-open when the approver does not respond; assuming an in-band kill-switch will work against a compromised agent; implementing a kill-switch that is never exercised.


References