eu-ai-act-mcp

Check GPAI Model Systemic Risk Classification

euaiact_check_gpai_systemic_risk

Read-onlyIdempotent

Check systemic-risk GPAI status under EU AI Act Art. 51 via training FLOPs threshold or Commission designation, and receive obligations under Arts. 52, 53, and 55.

Instructions

Determine whether a general-purpose AI model qualifies as a GPAI model with systemic risk under Art. 51. This tool uses an adjudicated conservative boundary at or above 10^25 cumulative training FLOPs for the Art. 51(2) presumption. The Commission may also designate models with equivalent capabilities or impact under Art. 51(1)(b). Returns baseline GPAI obligations under Art. 53 plus systemic-risk-only obligations under Art. 55, and the Art. 52 notification duty.

Input Schema

TableJSON Schema

Name	Required	Description
`model_name`	No	Optional model name for traceability in the response.
`training_flops`	No	Cumulative training compute in FLOPs (e.g. 2e25). The adjudicated product boundary triggers the Art. 51(2) presumption at or above 1e25. Omit if unknown; the tool then abstains instead of answering.
`commission_designated`	No	Whether the Commission has formally designated the model as GPAI with systemic risk under Art. 51(1)(b).

Output Schema

TableJSON Schema

Name	Required	Description	Default
`model_name`	Yes
`flops_threshold`	Yes
`notification_duty`	Yes
`relevant_articles`	Yes
`crosses_flops_threshold`	Yes
`systemic_risk_designation`	Yes
`is_gpai_with_systemic_risk`	Yes
`baseline_obligations_art_53`	Yes
`systemic_risk_obligations_art_55`	Yes

Schema Changelog

Changes observed during successful MCP inspections.

Changed5 schema fields changedv1.5.0Aug 19, 2026

changedInput schema / properties / training_flops / description

Previous value: -"Cumulative training compute in FLOPs (e.g. 2e25). Art. 51(2) presumes systemic risk when > 1e25."New value: +"Cumulative training compute in FLOPs (e.g. 2e25). The adjudicated product boundary triggers the Art. 51(2) presumption at or above 1e25. Omit if unknown; the tool then abstains instead of answering."

addedInput schema / properties / training_flops / minimum
```
Added value: +0
```

changedOutput schema / properties / crosses_flops_threshold / type

Previous value: -"boolean"New value: +[
+  "boolean",
+  "null"
+]

changedOutput schema / properties / is_gpai_with_systemic_risk / type

Previous value: -"boolean"New value: +[
+  "boolean",
+  "null"
+]

changedOutput schema / properties / systemic_risk_designation / enum

Previous value: -[
-  "threshold_met",
-  "commission_designated",
-  "none"
-]New value: +[
+  "threshold_met",
+  "commission_designated",
+  "none",
+  "undetermined"
+]

First observedv1.1.5Jun 14, 2026

TDQS

A4.1/5.0

Behavior4/5

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

Annotations already provide readOnlyHint: true and idempotentHint: true. The description adds critical behavioral context: the specific 10^25 FLOPs threshold ('adjudicated conservative boundary'), that it 'returns' specific obligations (Art. 53, 55, 52 notification duty), and that it can abstain when training_flops is omitted. This meaningfully enriches the safety profile beyond what annotations declare.

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

Conciseness4/5

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

Three sentences, well-structured, front-loads the core purpose in sentence one, then details the threshold and return values. No filler. Loses a point for the potential 10^25 vs 1e25 notational discrepancy that could cost a reader a second to reconcile.

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 output schema exists, the description need not list return types. For a legal determination tool with three optional params, it covers the test logic, edge case (abstain on unknown), and obligations scoping well. The slight ambiguity around what happens with conflicting inputs (e.g., high FLOPs but commission_designated: false) keeps it from a 5.

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

Parameters3/5

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

Schema coverage is 100% with rich parameter descriptions already documenting the default abstain behavior on training_flops and the commission_designated flag semantics. The description adds the overall threshold (10^25) that complements but doesn't conflict with the schema's 1e25 mention—actually introducing a slight inconsistency: description says 'at or above 10^25' while schema says 'at or above 1e25' (which are the same, but described with different significant figures, potentially confusing). Moderate value added beyond schema, which is already strong.

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 uses a specific verb ('Determine') and names the exact resource (general-purpose AI model systemic risk classification under Art. 51). It clearly distinguishes from siblings like euaiact_assess_system (which assesses broader system risk) and names the specific Art. 52/53/55 obligations returned, making its scope unmistakable.

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 implies when to use this tool versus alternatives by specificity ('GPAI model with systemic risk under Art. 51') and names the sibling alternative implicitly through distinct scope (e.g., it's not about penalties or deadlines). It lacks explicit 'use instead of X' language but provides clear context for models versus systems, distinguishing it from euaiact_assess_system and euaiact_classify_system without naming them.

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

Deploy Server