AI transparency
Last updated: November 2026
We inform you about the use of AI in line with the transparency obligations of the EU Artificial Intelligence Act (Regulation (EU) 2024/1689).
- Risk scores and list matches are rule-based (name similarity, birth year, country, list category) — no machine-learning model decides about a person.
- Deep investigation (optional): if enabled by the operator, a large language model (Anthropic Claude) extracts facts such as roles and company links from public web pages. Every AI-extracted fact is kept only if its exact quote is found on the source page, and it is shown with that quote and link so you can check it.
- AI research agent (optional): if enabled, Claude searches the web for the screened person or company, its public roles, risk issues (sanctions, enforcement actions, fines, cases, adverse media) and named business relationships, and then follows the most important connections for up to two more steps. A fact is shown only if its link came back from the web search itself; quotes are checked again on the source page where the site allows it and are then marked "verified on page". Results are suggestions for a human reviewer — the AI does not decide about a customer (Art. 22 GDPR), and every finding must be confirmed in official registers before acting.
- AI is not used for emotion recognition, biometric identification or facial recognition, and images are shown only as links.
- Results always require human review (see the compliance notice).