Skip to main content
Glama

Classify EU AI Act risk

check_ai_act
Read-only

Classify a product's risk level under the EU AI Act — minimal, limited, high or prohibited — and return the applicable obligations, each with its compliance deadline (limited-risk transparency duties apply from Aug 2, 2026; Annex III high risk from Dec 2, 2027). The boolean parameters fall into three groups, all optional and defaulting to the safest 'not applicable' value: (1) transparency triggers — usesAI, interactsWithPeople, generatesContent, automatedDecisions; (2) the eight prohibited practices of art. 5 — socialScoring, realtimeBiometricPublic + realtimeBiometricLawEnforcement, emotionRecognitionWorkEducation, biometricCategorisationSensitive, untargetedFaceScraping, manipulativeOrExploitative, individualPredictivePolicing; (3) high-risk triggers — embeddedInRegulatedProduct, annexIII domains. Answer the boolean questions about what the system does; leave unknowns at their defaults, which never over-report risk. Read-only and instant.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
usesAINoDoes the product use AI at all (LLM calls, recommendations, computer vision…)? false ⇒ minimal risk, the AI Act does not apply.
annexIIINoAnnex III high-risk domains that apply. Only these exact values count; anything else is not Annex III and must be left out.
socialScoringNoDoes it score people's social behavior or traits causing detrimental treatment in unrelated contexts, or disproportionate to the behavior (art. 5.1.c)? Prohibited. A seller rating or fitness points app is NOT this.
generatesContentNoDoes it generate text, images, audio or video shown to users? Triggers the art. 50 duty to label AI-generated content.
automatedDecisionsNoDoes it make automated decisions with legal or similarly significant effects on people (credit, hiring, admissions…)? Adds GDPR art. 22 duties; high risk only if an Annex III domain applies.
interactsWithPeopleNoDo people interact directly with the AI (chatbot, voice assistant…)? Triggers the art. 50 duty to disclose they are talking to an AI.
untargetedFaceScrapingNoDoes it untargetedly scrape facial images from the internet or CCTV footage to build or expand a face-recognition database (art. 5.1.e)? Prohibited. Matching a user's own consented selfie against their own ID photo is NOT this.
realtimeBiometricPublicNoReal-time remote biometric identification in publicly accessible spaces (e.g. live face recognition)? High risk (Annex III biometrics) — prohibited only when combined with realtimeBiometricLawEnforcement.
embeddedInRegulatedProductNoIs the AI a safety component embedded in an Annex I regulated product (toys, machinery, medical devices…)? High risk via art. 6.1, deadline Aug 2, 2028.
manipulativeOrExploitativeNoDoes it use subliminal techniques beyond a person's consciousness, or exploit a known vulnerability (age, disability, specific social or economic situation) to materially distort behavior and cause harm — e.g. dark patterns targeting a diagnosed gambling addiction (art. 5.1.a-b)? Prohibited. Ordinary persuasive marketing or UX nudges aimed at the general population are NOT this.
individualPredictivePolicingNoPredicting an individual's criminal risk based solely on profiling or personality traits (art. 5.1.d)? Prohibited.
emotionRecognitionWorkEducationNoDoes it infer emotions (frustration, attention, mood…) from employees at work or students in education, e.g. to flag disengaged staff or bored students (art. 5.1.f)? Prohibited. Emotion inference on end users of a public consumer product (not their employees/students) is NOT this.
realtimeBiometricLawEnforcementNoOnly if realtimeBiometricPublic: is it used FOR LAW ENFORCEMENT purposes? That combination is prohibited under art. 5.1.h and already in force; any other use of public real-time biometric ID is high risk instead. Without this the prohibition can never be reached.
biometricCategorisationSensitiveNoDoes it use biometric data (face, voice, gait…) to infer sensitive attributes — race, political opinion, religion, trade-union membership, sexual orientation (art. 5.1.g)? Prohibited. Ordinary face-unlock or liveness checks that don't infer these attributes are NOT this.

TDQS

A4.6/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

The annotations already declare readOnlyHint=true and destructiveHint=false, and the description reinforces this with 'Read-only and instant.' It goes beyond annotations by adding crucial behavioral context: defaults 'never over-report risk' and specific compliance deadlines for obligations. This transparency about safety defaults and output behavior exceeds what annotations provide.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is front-loaded with the core purpose, then efficiently enumerates parameter groups in a numbered list. Every sentence earns its place: purpose, deadlines, parameter grouping, and usage guidance. Despite covering 14 parameters, it remains dense and structured with no fluff. This is a model of concise, high-density documentation.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool's complexity (14 optional parameters, no output schema), the description covers the essential aspects: what it returns (risk level, obligations with deadlines), how to answer parameters, and the conservative default behavior. It does not specify the exact structure of the returned obligations, but it gives sufficient high-level information for an agent to know what to expect. The lack of an output schema raises the burden, and a short note on the return format would make it a 5.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so baseline is 3. The description adds value by grouping the 14 booleans into three logical categories (transparency triggers, prohibited practices, high-risk triggers) and explaining the dependency between realtimeBiometricPublic and realtimeBiometricLawEnforcement. This organization aids the agent in understanding the parameter structure, going beyond the schema's per-parameter descriptions.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool's function: 'Classify a product's risk level under the EU AI Act — minimal, limited, high or prohibited — and return the applicable obligations, each with its compliance deadline.' This is a specific verb+resource with distinct output. It also distinguishes itself from sibling tools (e.g., check_website, check_store) by focusing on AI Act risk classification.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description provides clear context for when to use the tool: it classifies products under the EU AI Act. It gives behavioral guidance ('Answer the boolean questions about what the system does; leave unknowns at their defaults') and notes the tool never over-reports risk. However, it does not explicitly mention alternatives or exclusions, so it stops short of a 5.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.

TDQS

A4.5/5.0
Disambiguation5/5

Each tool has a clearly distinct purpose: check_ai_act classifies AI risk, check_store and check_website audit different platform listings, claim_app creates an app, generate_policies drafts legal docs, get_claim_status tracks claim confirmation, and get_install_snippet provides installation code. No two tools overlap in function.

Naming Consistency5/5

All tool names follow a consistent verb_noun pattern with snake_case (e.g., check_ai_act, claim_app, generate_policies). The naming is predictable and easy to understand.

Tool Count5/5

Seven tools is a well-scoped set for the server's purpose: three for compliance checks, one for app creation, one for policy generation, one for status tracking, and one for installation snippets. Each tool earns its place without being excessive.

Completeness4/5

The tool surface covers the primary workflows: checking compliance, creating an app, generating policies, and installation. A minor gap is the lack of update/delete tools for apps, but the core lifecycle is covered. Agents can successfully complete the intended tasks.