The AI Act is coming. TalkWithData ships risk classification, transparency reports, human-oversight flows and model documentation — built in.
GPAI rules (Aug 2025), high-risk requirements (Aug 2026), full effect (Aug 2027). Waiting means scrambling under deadline.
Is your HR bot high-risk (Annex III)? Your AI support bot? Definitions unclear, legal advice needed per use case.
Art. 13: users must know they're talking to AI. Art. 50: AI-generated content must be recognisable. Not solvable ad hoc.
Per bot a classification + checklist per Annex III. What to document, monitor, report.
"This answer was generated by AI based on [sources]" — required under Art. 13/50, baked into every bot.
For high-risk decisions: human-in-the-loop confirmation, escalation paths, audit logs of who approved what.
Helper tool to classify your bot: minimal risk, limited transparency, high-risk Annex III, or prohibited. With legal source.
Datasheet per bot: training data (which knowledge base), model provider, evaluation results. Per Art. 11 requirements.
Every answer traceable to source + LLM call. Required for high-risk; useful for all bots.
Classification scan free. AI Act readiness report in a week.