Legal

AI Use & Data Processing

Last updated: 19 November 2026

This page explains how Aliaa uses artificial intelligence, what your organisation accepts when it uses it, and what happens to the data involved. It is part of the Terms of Service (the "AI Terms"). Words defined in the Terms mean the same here.

1. What the AI does#

Aliaa's AI features run on Claude models made by Anthropic PBC, used through Anthropic's commercial API. The AI:

  • answers questions in the chat and drafts texts (proposal sections, budgets, letters, funder reports);
  • reads a document you file, to record what it is about (type, topic, date, funder) so it can be found later;
  • summarises incoming mail in a connected mailbox and prepares a reply draft;
  • for organisations that give grants: summarises applications and grantee reports, runs an eligibility pre-check and flags points for reviewers;
  • drafts evaluation forms and groups their answers into themes;
  • searches the public web when a User asks it to research a question.

Some of these steps run when a User asks. Others run automatically when material arrives — for example when a file is filed or a mail arrives from a known contact.

If we change or add an AI provider, we will update this page and the sub-processor list in the DPA first, with the notice the DPA describes.

2. The rule: AI drafts, people decide#

  • The AI cannot send email. It has no sending ability at all. With a mailbox connected it can prepare a draft; a message goes out only when a person approves it and presses send.
  • The AI does not change your record by itself. What it proposes is saved as a draft and takes effect only when a person confirms it.
  • The AI sees only what the person using it is allowed to see. Material marked restricted — HR files, participant lists and case notes — is kept from the AI by default.
  • Uploaded files need a tick. A file a person uploads for drafting is read by the AI only after the uploader ticks that it holds no participant names, ID numbers or case notes.
  • The AI does not make decisions about people. It does not award or refuse grants or choose beneficiaries. It summarises and flags; your people decide.

3. What your organisation acknowledges#

By using the AI features, your organisation acknowledges and agrees that:

  1. Aliaa uses AI. The chat and drafting tools are an AI system, not a person. Your organisation will tell its Users so.
  2. AI Output can be wrong. It may contain errors, made-up facts, figures, dates, quotes or sources, wrong calculations, outdated information, or text that does not fit a funder's requirements. It can sound certain when it is not.
  3. Human review is required. A competent person must check every AI Output — especially numbers, budgets, dates, eligibility statements and anything about named people — before it is relied on, approved or shared.
  4. Your organisation is responsible for what it sends. Anything approved and sent to a funder, applicant, grantee, authority or anyone else is your organisation's own communication, whoever drafted it. We are not liable for loss caused by AI Output that your Users approved, sent or relied on (Terms, sections 6 and 15).
  5. Laws and funder rules on AI are yours to follow. Some funders restrict AI use in applications and reports or require you to disclose it. Checking and following those rules is your responsibility; we do not warrant that your use of the AI features meets them (Terms, section 14.2).
  6. AI Output is not professional advice — legal, tax, financial, accounting or other.
  7. The saved record prevails. Your record holds what a person confirmed. If AI Output differs from that record or from the original document, the record and the original control.

4. What data the AI receives#

For each AI step, Aliaa sends Anthropic what that step needs: the User's question or instruction, the relevant parts of your organisation's record, and the text of the document, mail or application being worked on. For web research, the User's question is sent and Anthropic's web search runs it.

In the automatic steps on incoming mail and on grant applications, Aliaa is built to strip email addresses, phone numbers and long number sequences (such as ID or bank numbers) before the text goes to the AI.

Reading the text out of your files (including scanned files) and indexing them for search happen on our own servers. Your files are not sent to any outside service for that.

Please do not give the AI highly sensitive personal data (for example health, criminal-record, biometric or ID-number data about beneficiaries or applicants) unless the task truly needs it. Never type passwords, access codes or card numbers into the chat or put them in uploaded files (Acceptable Use Policy, item 16).

5. No training on your data#

  • We do not use your data, your requests or AI Output to train AI models.
  • Anthropic does not train on it either. Anthropic's Commercial Terms state that it "may not train models on Customer Content from Services".
  • We do not sell your data or use it for advertising.

6. Where the data goes and how long it stays#

Who Where What is kept
Your Instance iDigital.Agency See the hosting location in the DPA, Annex 3 Your conversations, drafts and AI Output stay in your own Instance until you delete them or the subscription ends.
AI processing Anthropic PBC United States Under Anthropic's published policy, API inputs and outputs are deleted within 30 days. Content that Anthropic's safety systems flag as breaking its Usage Policy may be kept for up to 2 years (and its safety scores for up to 7 years), and data may be kept longer where the law requires.

Anthropic acts as our sub-processor under its Data Processing Addendum, which includes the European Commission's Standard Contractual Clauses. We will update this section if Anthropic changes these terms.

7. Costs and limits#

AI usage is metered and billed as set in your Order. Each organisation has a daily and a monthly AI spending limit. When a limit is reached, the AI features pause until the next period or until an authorised person raises the limit. The rest of Aliaa keeps working.

8. Using AI responsibly#

The Acceptable Use Policy applies to the AI features. In particular, do not use them to mislead funders or the public, to create false documents or invented evidence, or to decide about people without a person's real review.

Intended use. Aliaa is intended as an assistant that prepares drafts, summaries and checks for people to review and approve. It is not designed as a high-risk AI system under the EU AI Act (Regulation (EU) 2024/1689). Your organisation agrees to use it only for this intended use: never for a practice prohibited by Article 5 of that Act, and never as a high-risk system — for example, it will not let AI Output decide, without a person's real review, who receives a grant, a benefit, a job or a place in a programme.

9. Questions#

[email protected]