Key takeaways
- UK GDPR has no special AI exemption: the same principles apply to a phone assistant, a chat widget or a rota tool as to a paper booking diary.
- Allergy and dietary details are health data, so AI tools that hear or read them need extra care and a clear Article 9 condition.
- Since the Data (Use and Access) Act 2025, solely automated significant decisions must come with safeguards: information, a chance to respond, human review and a way to contest.
- A short DPIA before switching on a new AI tool is usually the quickest way to show the regulator you thought about risk.
Where AI already meets guest data in a venue
Most hospitality operators are not running AI projects. In late September 2026, 70.7% of UK accommodation and food businesses said they used no AI technology at all, and another 11.4% were not sure[1]. Yet the tools arrive anyway, bundled into booking platforms, review dashboards, phone systems and payroll software. Where AI is in use, 39.2% of these businesses said it was to improve operations and 17.8% to provide or personalise services for customers[2].
Source: Office for National Statistics, 2026[1]
Show the data
| Item | Value |
|---|---|
| No AI used | 70.7% |
| Not sure | 11.4% |
| Visual content AI | 10.3% |
| Text generation AI | 8.9% |
Guests are further ahead than venues. Six in ten UK adults had used an AI chatbot in the previous three months when the government last tracked it[3], but public associations with AI are still mostly negative[4]. That gap is the practical reason to get data protection right: a guest who feels tricked by a bot is a guest who complains, and complaints are how most organisations first meet the Information Commissioner's Office.
- AI phone answering that records calls, transcribes them and writes summaries.
- Website chat that answers menu, allergen and opening-time questions and collects booking requests.
- Review-reply drafting tools that read guest reviews, some of which name staff or mention illness.
- Forecasting and rota tools that use staff availability, sickness and performance data.
- Recruitment tools that rank or filter applicants for bar, floor and kitchen roles.
- CCTV or footfall analytics that count, track or identify people.
Pick a lawful basis for each use, not one for the whole tool
The ICO's starting point is the UK GDPR principles: lawful, fair and transparent processing, collected for a stated purpose, kept to what you need, accurate, held no longer than necessary and kept secure. An AI feature does not change those duties; it changes how easy it is to drift past them. Write down each purpose separately, because one product can do several jobs with different legal footing.
| AI use | Personal data involved | Likely lawful basis | Watch-out |
|---|---|---|---|
| Taking a booking request by phone | Name, number, party size, date, notes | Contract (steps before a booking) | Recording and transcripts need a retention period |
| Answering allergen questions in chat | Health details typed by the guest | Contract plus an Article 9 condition such as explicit consent | Hand the allergen decision to a trained person |
| Drafting review replies | Reviewer name, story, sometimes staff names | Legitimate interests | Do not paste health or staff details into public replies |
| Ranking job applicants | CV, availability, right-to-work answers | Legitimate interests or contract steps | Likely a significant automated decision; safeguards needed |
| Footfall or CCTV analytics | Images, movement patterns | Legitimate interests | Large-scale monitoring of public areas triggers a DPIA |
If you rely on legitimate interests, keep a short balancing note: what you gain, what the guest or worker would reasonably expect, and what you do to limit the intrusion. A paragraph per use is enough for a small venue.
Tell people, in the channel where it happens
Transparency is where hospitality most often slips. A privacy notice buried in the website footer does not help a caller who has just been greeted by an assistant. Put the key facts where the interaction starts and keep the full detail in your privacy notice: who you are, why you are collecting the details, how long you keep recordings and transcripts, who your suppliers are and how to object or complain.
Example phone greeting: 'Thanks for calling The Lantern. You're speaking to our AI assistant, which can take booking requests and answer questions. Calls are recorded and summarised so our team can follow up. Say “staff” at any time to reach a person.'
The same applies to chat. A one-line notice above the message box, a link to the privacy notice and a clear way to reach a person cover most of what a guest needs. If you use Ava or any other AI phone assistant, check what is stored after each call (Ava keeps a summary and transcript in the team inbox) and set a deletion period you can defend. For wording and timing, see telling callers they are speaking to AI.
Automated decisions after the 2025 changes
The Data (Use and Access) Act 2025 replaced the old Article 22 with new Articles 22A to 22D. A decision counts as solely automated where there is no meaningful human involvement, and it is significant if it has a legal or similarly significant effect on the person[5]. Where a significant decision is taken solely by automated means, you must have safeguards that give the person information about the decision, let them make representations, let them obtain human intervention and let them contest it[6].
Most guest-facing tools in a restaurant will not reach that bar. Suggesting a table time, answering a question about parking or drafting a reply is not a significant decision. Two areas deserve a closer look:
- Recruitment: a tool that automatically rejects applicants for a waiting job before anyone reads the application is the clearest hospitality example of a significant automated decision.
- Guest blocking: a system that silently refuses bookings or demands deposits from people it profiles as likely no-shows. Keep a manager in the loop and give the guest a route to ask why.
- Rota and performance scoring: if hours or shifts are cut by a score with no human review, treat it as significant and add the safeguards.
Meaningful human involvement means a person who can change the outcome and actually looks at the case, not a manager who clicks approve on a list of 200 rejections.
Allergy and dietary details are health data
About 23% of adults in England, Wales and Northern Ireland report a food hypersensitivity such as an allergy or intolerance[7], so a busy venue hears health information every service. Under UK GDPR, data concerning health is special category data and needs a condition under Article 9 as well as a lawful basis. In practice, for AI tools that means three rules:
- Collect only what the kitchen needs to serve the guest safely, such as the allergen and the severity, not a medical history.
- Do not let an AI tool give the final answer on whether a dish is safe. Route allergen questions to a trained member of staff and record the handover.
- Keep health notes out of marketing lists, review replies and any data sent to a supplier for training their models.
When a DPIA is expected
UK GDPR requires a data protection impact assessment where processing, in particular using new technologies, is likely to result in a high risk to people's rights and freedoms, and it must be done before the processing starts[8]. Article 35 names three cases where an assessment is in particular required: systematic and extensive automated evaluation of people that leads to significant decisions, large-scale processing of special category data, and systematic monitoring of a publicly accessible area on a large scale. A new AI phone or chat tool that records every caller, hears allergy details and runs on technology your team has not used before is the kind of change worth screening, even for a single site. For a step-by-step version sized for a venue, use our guide to data protection impact assessments for guest tech.
Security and suppliers
In 2025, 30% of UK food or hospitality businesses identified a cyber breach or attack in the previous 12 months[9]. AI tools add new places where data sits: call recordings, chat logs, prompts typed by staff. Ask every supplier where data is stored, whether it is used to train models, how fast they report a breach and how you delete records. Our list of questions to ask an AI phone provider turns these into a buying checklist.
A one-page compliance routine
List every AI feature
Include features inside tools you already pay for, such as review replies in your booking platform or smart summaries in your phone system.
Write one purpose and lawful basis per feature
Add an Article 9 condition anywhere health or allergy details can appear.
Screen for high risk
New technology, monitoring, health data at scale or automated decisions about people mean you complete a DPIA before go-live.
Update notices at the point of contact
Phone greeting, chat banner, job advert and privacy notice should all say what the tool does and how to reach a person.
Check the supplier contract
Confirm a processor agreement, storage location, training use, breach notice and deletion on exit.
Set retention
Pick a deletion period for recordings, transcripts and chat logs and check it is applied automatically.
Train the team
Staff should know what not to paste into AI tools and how to handle a guest who asks what data you hold. A written AI use policy helps.
Review every six months
Features change with software updates. Re-run the list and note what changed.
Frequently asked questions
Does UK GDPR ban restaurants from using AI?
No. UK GDPR applies to AI tools the same way it applies to any processing of personal data. You need a lawful basis, clear information for guests and staff, security, sensible retention and, for higher-risk uses, a DPIA.
Do I have to tell callers they are speaking to an AI assistant?
Being open about it is the safest reading of the transparency principle, and guests react better when told up front. A short line in the greeting, plus a way to reach a person, covers it.
Is an allergy note on a booking special category data?
Yes. Information about a person's allergy or intolerance is data concerning health. Collect only what the kitchen needs, restrict who can see it and keep it out of marketing.
What counts as a significant automated decision in hospitality?
The clearest examples are automatically rejecting job applicants or refusing a guest service based on a profile, with no meaningful human review. Answering questions or suggesting a table time is not.
References
Office for National Statistics (2026). Business insights and impact on the UK economy, wave 165 dataset (AI sheets) (opens in a new tab) ↑
Office for National Statistics (2026). Business insights and impact on the UK economy, wave 165 dataset (AI sheets) (opens in a new tab) ↑
Department for Science, Innovation and Technology (2024). Public attitudes to data and AI: Tracker survey (Wave 4) report (opens in a new tab) ↑
Department for Science, Innovation and Technology (2024). Public attitudes to data and AI: Tracker survey (Wave 4) report (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2025). UK GDPR, Article 22A: automated processing and significant decisions (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2025). UK GDPR, Article 22C: safeguards for automated decision-making (opens in a new tab) ↑
Food Standards Agency (2025). Food and You 2: wave 11 key findings (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2016). UK GDPR, Article 35: data protection impact assessment (opens in a new tab) ↑
Department for Science, Innovation and Technology (2025). Cyber security breaches survey 2025 (opens in a new tab) ↑