AI with people
in control

• Document check: reading an uploaded document to see which required document it is and which step it belongs to, and noting facts such as dates.
• Authority letters: reading a letter the team uploads (for example a request for more documents or a decision) to prefill the form the team then checks.
• Rule research: helping Settio’s own staff research official sources when they prepare country rules, which people review before publishing.
• Help answers: answering how-to questions about Settio from our help guides only.
• Free personal account: finding trips in a file or text the person sends, and reading the kind and dates of a document they put in their vault, within a daily limit.

• It never reads passports, national ID cards or residence permit cards in the platform or the portal.
• It never decides anything about a person: not eligibility, not an application, not a hire.
• It never scores or ranks candidates or employees.
• It never sends anything to an employee or an authority by itself.

• We use Anthropic’s Claude models. Anthropic is our processor, doesn’t use the data to train its models, and keeps it only for the limited time its commercial terms allow. It is in the United States; transfers rely on the EU Standard Contractual Clauses (see subprocessors).
• The document check is given only the case’s own list of required documents. Text inside a document is treated as data, never as an instruction.
• The personal account’s document reading is told to return no names or numbers, only a kind and dates.

• Every AI result is shown as a suggestion and marked as AI.
• A letter reading fills in a form; nothing is saved until a person confirms it.
• The team can correct any document match, and employees can always ask a person on their team.

AI can be wrong or incomplete. Treat every AI result as a suggestion, not as legal, immigration or tax advice, and check it against the document. See Information, not legal advice.

Settio’s AI features support people doing paperwork. They don’t make or rank decisions about workers, so they are not used as a high-risk system under the EU AI Act (Regulation (EU) 2024/1689, Annex III point 4). We tell people when a result comes from AI (Art. 50).