Compliance & regulation
AI Act (EU)
EU regulation adopted in 2024 that imposes risk-tiered obligations on AI systems deployed in the European Union, including the AI moderation engines used by comment platforms.
The EU AI Act is the regulation adopted by the European Union in 2024 that classifies AI systems by risk tier and imposes obligations accordingly. For news publishers running AI-powered content moderation, the most relevant provisions are :
- Risk-tier classification : moderation AI is typically classified as limited risk, requiring transparency to the user (the reader must know an AI is involved in the moderation decision). Combined with DSA Article 14 this is mostly already covered.
- General-purpose AI obligations : if the moderation engine relies on a third-party LLM (Mistral AI, GPT-4, etc.), the provider has transparency obligations that flow down to the deployer.
- Documentation : technical documentation of the AI system, including training data sources, model version, evaluation methods. Logora maintains this internally and shares on contractual demand.
- Human oversight : automated decisions affecting users must have a meaningful human review path. Logora’s 15% human queue is the operational answer.
What it changes for Logora
Logora’s moderation pipeline was already compliant by design pre-AI Act : transparent statement of reasons (DSA Article 14), structured logging, human review on borderline content, EU-hosted model. The AI Act formalised the obligations rather than introducing new ones.
For Mistral AI (used since April 2026 for debate summaries), the model provider documentation flows down to Logora and can be shared with your DPO.
See AI moderation, DSA, and GDPR for the surrounding regulatory frame.