DocuLiq — AI Notice

Version 1

In short

DocuLiq is not professional advice and can make mistakes. Check important information and track your legal deadlines yourself.

What is AI-generated

Summaries, risk reviews, extracted information, chat answers, translations, the descriptions of reminders found in a document and the tasks suggested from a risk review are produced by artificial intelligence (Google's Gemini models on Vertex AI, run in the European Union). Each of these results carries an AI label (AI-generated) in the app; the label is removed when you edit the result yourself. Every reminder DocuLiq finds automatically in a document carries this label, even when the AI did not rewrite it; a reminder you add yourself does not. Dates are found in the text by fixed rules, not by AI.

Text recognition (reading the text of scanned pages and images) and audio transcription are automated steps too. They can misread or mishear: a wrong letter, number or date in the extracted text carries over into everything built on it.

What the review does

• It scans the text of your document and lists the dates it finds (by rules) and the obligations and possible risks it finds (with AI).
• It can miss something, and a finding can be incomplete or wrong.
• The risk level (none, low, medium or high) and the score next to it are the AI's assessment of the document, not a calibrated measurement. Sorting documents by risk uses this assessment. If a workspace owner or admin changes the level or score by hand, that value is used instead.
• The tasks and the suggestions for the parties that a risk review proposes are general points to check. They are not binding and not a legal opinion.
• Reminders found in a document are added to the workspace's reminders automatically and are delivered unless you dismiss them. The AI may rewrite a reminder's description, and it may mark a date as not being a deadline; such a date is not delivered. You can edit or dismiss any reminder.

What Google does with AI inputs

Google states in its Vertex AI documentation that it does not use content sent to its Gemini models to train or fine-tune its models without the customer's permission or instruction. Google may keep inputs and results in memory for up to 24 hours to answer faster. If Google's automated safety systems flag a request as possible abuse of its service, Google may log that request for up to 90 days, in the same region, and authorised Google staff may review it.

Reporting an AI result

You can rate any AI result from the app with the flag next to it (Give feedback on this answer): helpful, wrong or misleading, or offensive or harmful. The feedback — the option you choose, the screen it came from, the document's identifier, and your name, email address and account identifier — is sent to our support mailbox, as the Privacy Notice explains. The text of the result is never sent.

What you should do

Check important findings against the document itself. For a decision with legal, financial or medical consequences, rely on the document and, where needed, on a professional.

Data controller: KIRIMLI TEKNOLOJİ LİMİTED ŞİRKETİ, Yakuplu Mah. Hürriyet Bul. Skyport Residence NO: 1 Kat: 10 İç Kapı NO: 113 BEYLİKDÜZÜ / İSTANBUL, Türkiye. Contact: support@doculiq.com.