Choose a London AI workshop you can complete with synthetic material, clear account rules and no need to upload client, employer or participant information.
Direct answer: Book the workshop only if its core exercise can be completed with organiser-supplied synthetic material or a safe file you prepare yourself. Before paying, confirm the required account, connected services, retention terms and whether observation is possible if a tool fails. If the learning outcome depends on uploading real work information, choose another session or ask for a redesigned exercise.
A trainer may say that participants should bring a “real problem” when they mean a realistic task. Those are not the same. You can test document classification, summarisation or extraction with invented content while keeping client, employee and unpublished business information out of the service.
The decisive question is not whether an event is described as private. It is whether you understand what information will leave your device, which account will process it, who can access it and what happens afterwards.
Applies to: in-person and hybrid London workshops where attendees use generative AI applications. UK data protection requirements may apply when personal information is processed; contractual, professional and sector duties can impose additional restrictions.
Use the zero-data workshop screen
The zero-data workshop screen is an editorial method for deciding whether a session remains useful when real work data is unavailable. “Zero-data” here means zero client, employer, participant or other operational information supplied by you. It does not mean the workshop platform performs no processing.
Assess four things before booking:
- the learning task you will complete;
- the material used to complete it;
- the account, integrations and permissions required; and
- the evidence you can take away and reproduce safely.
This applies the task-first approach in How to Choose the Right AI Tool Without Being Sold by the Demo. The workshop should help you test one defined job, not reward the most revealing upload.
Define a useful outcome before reading the event page
Write one sentence describing what you need to learn. For example: “I need to compare a long source with a summary and identify omitted exceptions.” That can be practised with a public policy, a synthetic contract or a fictional project brief. “I want to use my client folder with AI” is not a learning outcome and grants the tool an unnecessarily broad role.
List the fields a representative example needs. A synthetic customer table might need an invented identifier, request type, date and resolution. It does not need copied names, email addresses, order histories or free-text complaints. Keep the invented data visibly labelled so nobody later mistakes it for an authentic record.
Check your organisation's policy and contract before taking work material to a London venue. A ticket, confidentiality notice or trainer assurance cannot grant permission that you do not have. If the task touches regulated or professionally privileged material, ask the appropriate internal owner or qualified adviser rather than relying on a general workshop rule.
Ask the organiser five specific questions
Ask in writing so you can compare the answer with the event description:
- Does the organiser provide synthetic files that support every core exercise?
- Must attendees create a personal account, start a trial or enter payment details?
- Will the trainer, organiser or other participants be able to see prompts or outputs?
- Are any files, prompts, recordings or workspace contents retained after the session?
- Can an attendee follow the teaching without connecting cloud storage or uploading a private file?
An answer such as “the tool is secure” is too broad. Ask for the relevant provider documentation and the exact account tier used. Consumer, team, enterprise and API access may have different controls and terms. Disabling model training, where offered, is not the same as immediate deletion or an assurance that nobody else can access the workspace.
The ICO's current AI guidance says the appropriate security measures depend on the processing and that personal information should be limited to what is necessary. The guidance is under review following changes to UK data protection law, so check its status again near publication or use. ICO guidance on AI security and data minimisation.
Compare the available workshop formats
| Workshop arrangement | Data exposure | Learning value | Default decision |
|---|---|---|---|
| Organiser supplies clearly synthetic files | Limited to the supplied exercise and account activity | High if the files represent the stated task | Usually suitable after account and retention checks |
| You create a synthetic example in advance | You control the example, but the service still processes prompts and outputs | High when the example preserves the task's difficult features | Suitable if policy and account terms allow it |
| You upload a redacted work file | Residual identifiers, context and confidential facts may remain | Potentially realistic, but the exposure is harder to assess | Do not assume redaction makes it permissible |
| You connect a live work drive or mailbox | Broad permission and secondary data may be exposed | More realistic, but disproportionate for most training | Decline unless formally authorised and technically controlled |
Local processing can reduce transmission to a third party, but it does not automatically create complete privacy. Files may remain in logs, temporary storage, backups or exported results. The device itself may be visible or shared at the event. Ask what the exercise actually does rather than treating “local” as a complete answer.
Calculate the value of a safe exercise
Consider an illustrative London workshop with an £85 ticket and £14.20 return travel. A freelance attendee expects to spend four hours in the session and thirty minutes travelling beyond their normal working day. They value that time at £24 per hour for planning purposes.
The direct cash outlay is:
£85 + £14.20 = £99.20.
The illustrative value of 4.5 hours is:
4.5 × £24 = £108.
The full decision cost is therefore £99.20 + £108 = £207.20. The £108 is not automatically cash lost. It becomes forgone income only if the person turns down billable work; otherwise it is a way to compare uses of time.
Suppose the agenda contains twelve exercises. Ten work with synthetic files, while two require a connected work account. The attendee should not treat those two as usable. The planning cost per safe exercise is £207.20 ÷ 10 = £20.72, before any follow-up practice.
That figure does not prove the workshop is worth £20.72 per exercise. It makes the trade-off visible. If the two unavailable exercises contain the only skill the attendee needs, the event is a poor fit. If one safe exercise prevents repeated manual rework, it may justify the cost. Compare the result with a free talk, official tutorial or a smaller internal trial rather than assuming paid attendance is necessary.
Prepare a synthetic file that preserves the difficulty
Remove real content first, then reconstruct the features that make the task difficult. For a summary test, include conflicting dates, an exception, a negative statement and a table. For extraction, include an empty field, an unusual unit and two similar headings. For classification, include borderline examples rather than only obvious cases.
Use invented organisations and people. Do not transform a real record by changing only its name. The remaining dates, amounts, circumstances and wording may still reveal the source or remain confidential.
Open the file before leaving for the venue and confirm that it contains no comments, revision history, hidden sheets, document properties or links back to work systems. Save a clean copy in a workshop-only folder. Keep real work accounts signed out if the exercise does not require them.
The NCSC's secure AI development guidance treats prompts, data, logs and assessments as assets that need appropriate protection. Its audience is mainly organisations developing systems, but the asset distinction is also useful when deciding what to take into a training room. NCSC secure AI development guidance.
Decide within 20 minutes before booking
- Spend five minutes defining the task and the minimum features a synthetic example needs.
- Spend ten minutes checking the event requirements and obtaining written answers to any missing account, visibility or retention questions.
- Spend five minutes comparing the complete ticket, travel and time cost with a safe alternative.
- Book only when the central outcome remains achievable without work data, and keep the organiser's answer with your booking record.
Stop if the organiser requires a live work connection, cannot explain who sees submitted material or dismisses your data concern without offering an alternative. Missing one London workshop is less costly than creating an exposure you were not authorised to create.
Related guides
Frequently asked questions
Is removing names enough to make a work document safe for an AI workshop?
No. A document can remain identifiable or confidential through dates, amounts, job titles, unusual events, account numbers, comments and combinations of facts. Removing direct identifiers is not the same as anonymisation, and anonymisation itself requires careful assessment. Start with a newly invented file rather than altering a real one when the learning task allows it. If authentic information is genuinely necessary, obtain the required organisational approval and assess the exact service, account and purpose. Do not let a trainer's informal use of the word “anonymous” replace your employer's data-handling requirements.
Can I use a personal AI account with synthetic workshop data?
Possibly, but first check whether the workshop and your organisation permit that account arrangement. Even synthetic content creates account activity, prompts and outputs, and a free trial may require payment information or renew automatically. Check the current provider terms, retention controls and export options for the exact plan. Keep work identities and integrations separate unless there is an approved reason to connect them. A personal account is not automatically private, while a business account is not automatically suitable for every dataset. The safe choice depends on the task, permissions and documented controls.
What if the trainer says the provider does not train on my data?
Treat that as one narrow claim to verify, not a complete privacy assessment. You still need to know what is retained, who can access it, which account tier the statement covers, whether connected tools receive it and how deletion works. Ask for the current official documentation and check the applicable region and plan. Even a confirmed no-training setting does not grant permission to upload confidential or personal information. Your organisation may prohibit the transfer for contractual, security or professional reasons that have nothing to do with model training.
Is an offline AI workshop always safer?
No. Offline processing can avoid sending material to a remote service, but safety still depends on the device, software source, permissions, logs and files produced. A workshop laptop may be shared, unmanaged or reset in an unclear way. A downloaded model or tool can also have licence and compatibility requirements. Use synthetic data unless a formally approved process establishes otherwise. Ask how the software was obtained, what it writes to disk and how workshop devices are cleared. “Offline” describes connectivity at one point; it does not prove complete privacy or security.
Should my employer pay if I cannot use company data in the session?
Only if the workshop still serves an approved work objective and the full cost compares well with alternatives. Real company data is not necessary for many useful exercises. Present the learning outcome, synthetic practice plan, ticket and travel cost, time required and intended follow-up test. The employer can then decide whether the session supports your role. Do not promise a measured return before attending. Afterward, report what you can reproduce safely and what remains unverified. A request for payment should not be framed as permission to connect company systems during the event.
May I photograph the exercise instructions instead of uploading anything?
Check the organiser's recording and copyright rules first, and avoid capturing attendees, badges, private screens or confidential examples. Permission to attend does not automatically permit photography or republication. A handwritten note of the task, required settings and source link may be enough. If photographs are allowed, frame only the authorised material and store it according to its purpose. Rights and privacy requirements vary by event, contract and jurisdiction. Ask the organiser or a qualified adviser when intended reuse goes beyond personal reference or the material contains other people's information.
Sources and verification
- ICO: security and data minimisation in AI, checked 6 October 2026 for the risk-based security and data-minimisation principles and the notice that the guidance is under review.
- NCSC: secure AI system development, checked 6 October 2026 for the treatment of data, prompts, logs and assessments as AI-related assets. It is used as organisational security guidance, not as a legal rule for attendees.
- The assigned parent guide was read from the supplied website source. Its exact internal URL is retained as instructed.
This article is practical guidance. Apply it in proportion to your tools, evidence, risks, and responsibilities.
