The EU AI Act's transparency rules (Article 50) have applied since 2 August 2026. This article explains how they relate to Foodback Intelligence and Foodback Copilot, which parts of Foodback use AI, and how responsibility is shared between Foodback and you.
ℹ️ This is general information, not legal advice. If you operate in the EU or serve EU guests, we recommend reviewing this with whoever handles compliance in your organisation.
Who is responsible for what
The AI Act splits responsibility between two roles:
Provider — the party that develops an AI system and places it on the market under its own name. For Foodback Intelligence and Foodback Copilot, that is Foodback. We build these features, decide how they work, and are responsible for how they are designed and disclosed.
Deployer — the party that uses an AI system under its own authority. When you use Foodback Intelligence in your dashboard, that is you.
Foodback Intelligence is built partly on third-party AI models, but the systems we put on the market are our own. We remain the provider, and we do not pass that responsibility on to the underlying model vendors.
Where AI is used in Foodback
Feature | What it does | Type |
Importance Rank | Ranks questions by influence on overall satisfaction and improvement potential | Statistical (regression analysis) — does not generate text |
Sentiment Analysis | Classifies written guest comments as Positive, Neutral, Mixed or Negative | Text classification — does not generate text |
Free Text Summary | Summarises free-text comments from Pulse, One More Question and Campaigns | Generative — produces AI-written text |
Recommended Actions | Suggests monthly actions based on your feedback trends | Generative — produces AI-written text |
Foodback Copilot | Answers questions about your data in natural language | Generative — produces AI-written text |
Everywhere AI is involved, you will see the Foodback Intelligence marker, and Foodback Copilot identifies itself as an AI assistant.
What Foodback does as provider
✅ We label AI clearly. Every dashboard element powered by AI carries the Foodback Intelligence marker, and Foodback Copilot identifies itself as an AI assistant.
✅ We are extending that marking beyond the dashboard. The AI Act requires AI-generated text to carry machine-readable marking, with a deadline of 2 December 2026 for features already in use. We are working to make AI-generated text remain identifiable when it leaves the dashboard — in Excel exports, copied text and scheduled email reports — as far as the current state of the art for text marking allows.
✅ We keep a human in the loop by design. Foodback Intelligence surfaces insight and suggestions. It does not act on your behalf, make decisions about individual people, or change anything in your account.
✅ We show our work. Summaries and sentiment are always traceable back to the underlying guest comments in All Foodbacks, so you can check any AI output against the source.
What Foodback does not do
These are the categories the AI Act treats most strictly, so it is worth being explicit:
❌ No biometric data. Foodback does not process facial images, voice recordings, fingerprints, or any other biometric data — at any point, in any product.
❌ No emotion recognition in the AI Act sense. Our Sentiment Analysis reads written text that guests choose to submit. Under the AI Act, an "emotion recognition system" is one that infers emotions from biometric data. The European Commission has confirmed that content and sentiment analysis of written text is not based on biometric data and falls outside that definition.
❌ No profiling of, or automated decisions about, individual guests. Foodback Intelligence works on aggregated feedback, not on individual people.
❌ No high-risk use. Foodback Intelligence is not designed or intended for recruitment, employee evaluation, workplace monitoring, education, credit scoring, or any other high-risk purpose listed in the AI Act.
What you should consider as deployer
Do not use it to evaluate or monitor individual employees. Foodback Intelligence is built to improve guest experience, not to assess staff. Using guest feedback to evaluate, rank or monitor named employees is your own decision, and may bring obligations on you under the AI Act as well as employment and data protection law.
Review AI output before you republish it. If you share a Free Text Summary, Recommended Action or Copilot answer outside your organisation — in a public report, a newsletter, or marketing material — you are responsible for checking it, and for disclosing that it was AI-generated where that applies to you.
Keep a human in the decision. Treat AI output as input to a decision, not the decision itself. This is good practice generally, and it keeps you clear of obligations that attach to automated decision-making.
🧠 On accuracy
AI can be wrong. Summaries, sentiment classifications and recommended actions are generated from your own data and should be treated as a fast starting point rather than a final answer. Where something matters, check it against the original guest comments, which are always available in All Foodbacks.
Questions?
If you need more detail for your own compliance work — including for a procurement or vendor assessment — reach out to the Foodback Support Team through the help center and we will point you to the right documentation.
For a full description of the features referenced here, see Foodback Intelligence.