Home / AGILE-AI
Private Sector
Stage 1, Stage 2

AGILE-AI

Compliance Tech Sociedade Unipessoal LDA (Complear)
‹ Back to Open Call Winners
AGILE-AI brings Complear OS, an agentic compliance infrastructure, to the AI-on-Demand (AIoD) ecosystem. Complear OS keeps regulatory documentation synchronized with the software it describes: it detects code changes that affect regulatory status, redrafts the affected technical documentation, risk files and DPIAs, and requires a named reviewer's cryptographic signature before any change enters the audit log. In Stage 2, Complear integrates three AIoD components (TAI Scan Tool, AI Testing Framework and the MeluXina GPU Starter Plan) and publishes an open-source Compliance Artifact Collector and metadata standard, so any AIoD asset can declare its conformity-readiness.

Background

Providers of high-risk and medical AI must document compliance with the EU AI Act, MDR/IVDR, GDPR and related standards, and keep that documentation current at every release. In most SMEs this is still done by hand, late in development, which delays market entry. Deployers such as hospitals hesitate to adopt third-party AI without evidence they can trust, as they carry liability for its use. AIoD offers AI assets, compute and trustworthy-AI assessment tools, but no step that turns an assessment into regulatory evidence a notified body or hospital will accept.

Proposed solution

Complear OS is an agentic compliance infrastructure. It connects to a provider’s code host and issue tracker, detects changes that affect regulatory status, and redrafts the affected sections of the technical documentation, risk management file and DPIA. Each draft goes to a named reviewer, whose cryptographic signature is required before it enters the audit log, in line with human oversight under Article 14 of the AI Act.

The AI approach is neuro-symbolic. A regulatory ontology encodes obligations clause by clause (EU AI Act, MDR/IVDR, GDPR, ISO 13485, ISO 14971, EN 81001-5-1, ISO/IEC 42001). Domain-tuned small language models on open-weight bases retrieve from the ontology and draft; a deterministic layer checks clause coverage and evidence traceability before human review.

The Compliance Artifact Collector, SDK, Conformity Metadata Standard and compliance test suites are released as open source under Apache 2.0. Source code stays inside the customer’s network; only signed evidence leaves it.

Objectives

  • Integrate the TAI Scan Tool as the intake step for AI Act risk classification and obligations
  • Build compliance test workflows on the AI Testing Framework measuring clause coverage, citation precision and unsupported statements
  • Adapt and evaluate domain-tuned language models on the MeluXina GPU Starter Plan
  • Publish the open-source Compliance Artifact Collector, SDK and Conformity Metadata Standard on AIoD
  • Validate continuous MDR and AI Act documentation on one clinical AI use case, preparing a Stage 3 pilot with an AI Adopter

Compliance Tech Sociedade Unipessoal LDA (Complear)

Portugal
Complear combines an AI-powered compliance operating system with boutique advisory to automate the regulatory lifecycle for safety-critical industries. Originating in Digital Health (EU MDR/IVDR, FDA) and expanding into Defence, Space, and Energy, it integrates AI, cybersecurity, and privacy frameworks to accelerate time-to-market.
Stay informed on AI opportunities
Get updates on new solutions, market trends, and adoption strategies delivered to your inbox.
The DeployAI project has received funding from the Digital Europe Programme (DIGITAL) under grant agreement №. 101146490 - DIGITAL-2022-CLOUD-AI-B-03
Stay informed on AI opportunities
Get updates on new solutions, market trends, and adoption strategies delivered to your inbox.
The DeployAI project has received funding from the Digital Europe Programme (DIGITAL) under grant agreement №. 101146490 - DIGITAL-2022-CLOUD-AI-B-03