Originally posted by @shotwellj on 2026-03-31T04:18:30Z
Migrated from superlinked/superlinked#164
Is your feature request related to a problem? Please describe.
With the EU AI Act enforcement deadline on August 2, 2026, AI-powered search and recommendation systems like Superlinked will need to demonstrate compliance with Articles 9-15 for high-risk AI systems. As a framework that handles data ingestion, embedding, and retrieval, Superlinked touches several compliance-relevant areas.
Describe the solution you'd like
Native EU AI Act compliance checks built into Superlinked's pipeline:
- Art. 10 (Data Governance): Input validation and schema enforcement for ingested data, PII handling before embedding
- Art. 11 (Documentation): Auto-generated documentation for search pipeline configurations
- Art. 12 (Record-Keeping): Structured logging of queries, retrieval results, and relevance scores
- Art. 14 (Human Oversight): Query rate limiting, result filtering controls
- Art. 15 (Security): Input sanitization for search queries, adversarial robustness for embeddings
Describe alternatives you've considered
I ran Superlinked through AIR Blackbox, an open-source EU AI Act compliance scanner (Apache 2.0). You can run it yourselves:
pip install air-blackbox
air-blackbox comply --scan . --no-llm --format table --verbose
Everything runs locally, no data leaves your machine.
However, having compliance patterns built natively into the framework would be more valuable than an external scanner, since it could enforce compliance at the pipeline level rather than just detecting gaps after the fact.
Additional context
Enterprise teams using Superlinked for search and recommendations in EU markets will need to demonstrate that their AI-powered retrieval systems meet the Act's requirements. The EU AI Act carries penalties of up to €35M or 7% of global turnover.
Originally posted by @shotwellj on 2026-03-31T04:18:30Z
Migrated from superlinked/superlinked#164
Is your feature request related to a problem? Please describe.
With the EU AI Act enforcement deadline on August 2, 2026, AI-powered search and recommendation systems like Superlinked will need to demonstrate compliance with Articles 9-15 for high-risk AI systems. As a framework that handles data ingestion, embedding, and retrieval, Superlinked touches several compliance-relevant areas.
Describe the solution you'd like
Native EU AI Act compliance checks built into Superlinked's pipeline:
Describe alternatives you've considered
I ran Superlinked through AIR Blackbox, an open-source EU AI Act compliance scanner (Apache 2.0). You can run it yourselves:
Everything runs locally, no data leaves your machine.
However, having compliance patterns built natively into the framework would be more valuable than an external scanner, since it could enforce compliance at the pipeline level rather than just detecting gaps after the fact.
Additional context
Enterprise teams using Superlinked for search and recommendations in EU markets will need to demonstrate that their AI-powered retrieval systems meet the Act's requirements. The EU AI Act carries penalties of up to €35M or 7% of global turnover.