EU AI Act high-risk obligations: what crowdfunding platforms must do before August 2027

Crowdfunding sits in a category the EU AI Act, Regulation (EU) 2024/1689, treats as high-risk. The obligations do not all land at once, and platforms that wait for the final deadline will be retrofitting governance onto systems that were never built for it.

The dates that matter for AI used in regulated crowdfunding workflows.

The deadlines that matter

Two dates shape the work:

• 2 August 2026, transparency. Where people interact with or rely on AI output, they must be told. In practice, a clear on-screen label that findings are AI-assisted and should be verified by a professional.
• 2 August 2027, the full high-risk obligations: logging, human oversight, accuracy, risk management and technical documentation.

The article-level obligations

• Article 12, Logging. Automatic, timestamped records of inputs, outputs and decisions, exportable on demand.
• Article 13, Transparency. Outputs interpretable by the people relying on them; sources and confidence visible.
• Article 14, Human oversight. A person can understand, oversee and override the system. It never decides autonomously.
• Risk management and accuracy. A documented process, monitored metrics, and a defined response when something goes wrong.

What to build now

None of this requires waiting. The governance approach is buildable today, and building it now is cheaper than retrofitting it later:

• Human sign-off as a hard step in the workflow, not an optional review.
• Immutable, exportable evidence records and audit logs from day one.
• A human-verified knowledge base, so the system cannot invent a citation.
• Continuous accuracy monitoring against a qualified human baseline.
• A critical-incident halt for any hallucinated regulatory reference.

Read how we govern Dacxi Verify →

This briefing is general information, not legal advice.

Next briefing

What a complete ECSPR deal pack actually contains

→