How UK engineering firms can monetise their knowledge as AI training data
Every engineering consultancy has decades of judgement locked in its files: design calculations, options appraisals, inspection reports, value-engineering notes, the reasoning behind why a scheme was changed at stage 4. Most of it is archived after the project closes and never earns anything again.
AI developers increasingly want exactly this kind of material. General-purpose models are trained heavily on public web text, but most of the checked, professional reasoning that chartered engineers produce every day sits in private project files rather than on the open web. This guide explains, in plain English, how a UK engineering firm can turn that know-how into licensed AI training data, what is actually valuable, and what to get right on data protection and intellectual property before anything leaves the building.
Why AI labs will pay for professional data
Licensing content for AI training is no longer hypothetical. In February 2024 Reuters reported that Reddit had agreed a deal letting Google train AI models on its content, worth about $60 million a year according to one source. Reddit's IPO filing later disclosed $203 million in AI data licensing contracts over two to three years.
Those deals were for forum posts, whereas engineering material is specialist, structured, and shows how qualified people reason under real constraints such as codes, budgets, ground conditions and risk. That is hard to find on the open web, which is why it is worth licensing rather than scraping.
The UK policy backdrop also points towards licensing. In March 2026 the government published its Report on Copyright and Artificial Intelligence, confirming that a broad text-and-data-mining exception with opt-out is no longer its preferred approach, and that it will keep monitoring the licensing market. For now, existing copyright law stands, so permission (a licence) remains the clean route.
What engineering knowledge is valuable (and what isn't)
Not everything in your document management system is useful. AI developers are generally looking for material that shows reasoning, not just outputs.
Typically valuable
- Design rationale and options appraisals: why option B beat option A, and what trade-offs were weighed.
- Worked calculations with commentary: not just the numbers, but the assumptions, load cases and checks.
- Inspection and condition reports: observations, diagnosis and recommended remedial work.
- Review and checking comments: a senior engineer's mark-ups on a junior's design are an excellent record of expert correction.
- Technical queries and responses (RFIs): real problems and how they were resolved.
- Lessons-learned and post-project reviews.
Usually less valuable, or harder to use
- Drawings with no accompanying explanation.
- Commercial correspondence, fee proposals and invoices.
- Material where the client owns the copyright, or which is covered by confidentiality obligations you cannot clear.
- Anything that cannot be properly anonymised.
Step 1: Check who owns the IP
Before thinking about value, establish that you can license the material at all.
Under the Copyright, Designs and Patents Act 1988, section 11(2), where a work is made by an employee in the course of employment, the employer is normally the first owner of the copyright, subject to any agreement to the contrary. So reports and calculations written by your staff will often belong to the firm.
However, your client appointments matter. Many consultancy agreements assign copyright to the client, grant the client a broad licence, or impose confidentiality obligations. Review your standard terms and any bespoke appointments, and take legal advice where they are unclear. A reputable data partner should agree IP terms with you in writing and never ask you to share material you do not have the right to license.
Step 2: Anonymise properly under UK GDPR
Engineering documents contain personal data more often than people expect: names of engineers and client contacts, signatures, site addresses linked to individual homeowners, photos with people in them.
The Information Commissioner's Office published anonymisation guidance in March 2025. Key points for firms:
- Data protection law does not apply to genuinely anonymous information, but the bar is "effective anonymisation": the risk of identifying someone must be sufficiently remote.
- The ICO's two key indicators are singling out and linkability, and you must consider means "reasonably likely" to be used by anyone who might get access, including a "motivated intruder".
- Pseudonymised data (for example, names swapped for codes where a key still exists) is still personal data in the hands of whoever holds the key.
In practice that means removing names, signatures, contact details and individual addresses, generalising locations where needed, and checking that a project cannot be re-identified from distinctive details combined with public information (planning portals, press releases, your own case studies).
Client confidentiality is a separate question from GDPR: even fully anonymised content may still be confidential to a client, so step 1 still applies.
Step 3: Decide what you are comfortable sharing
You should stay in control. A sensible process looks like this:
- Scope: pick a discipline, era or project type to start with, such as bridge inspection reports or drainage design notes.
- Sample: prepare a small, anonymised sample so the data partner can assess quality and fit.
- Approve: review the structured, anonymised output before release. Nothing should leave without sign-off.
- License: agree written terms covering permitted use, exclusivity, duration and payment.
Step 4: Understand the commercial model
There are two broad ways to be paid.
One-off licence fee
A fixed payment for a defined dataset. Simple and predictable, but you get nothing more if the data proves valuable to several buyers.
Revenue share
You receive a share of what the dataset earns each time it is licensed. Both sides gain from more sales: your partner has a reason to find more buyers, and you benefit from repeat licensing of the same material. It works best where the partner handles the structuring, anonymisation and sales, so your team's time stays on fee-earning work.
Whichever model you choose, insist on written clarity about who the licensees can be, what they may do with the data, and how revenue is reported to you.
Practical checklist for engineering firms
- Identify a document set rich in reasoning (reports, calcs with commentary, checking notes).
- Review client appointments for copyright and confidentiality terms.
- Confirm the firm owns, or can license, the material.
- Anonymise in line with ICO guidance, and test for re-identification.
- Approve every dataset before it is released.
- Get IP, permitted-use and payment terms in writing.
- Take independent legal advice where you are unsure.
FAQ
Is it legal for a UK engineering firm to license its documents for AI training?
Generally yes, where the firm owns the copyright (or has the right to license it), client confidentiality obligations allow it, and personal data is either removed through effective anonymisation or handled lawfully under UK GDPR. Check your appointments and take legal advice on edge cases.
Will our clients' projects be identifiable?
They should not be. Good practice is to remove names, addresses and distinctive identifiers, and to test whether a project could be linked back using public information. You should approve the anonymised output before anything is shared.
How much time does it take our team?
It depends on the partner. With a managed service, your main tasks are choosing what to share, checking IP and confidentiality, and approving the final dataset.
What about older archives from completed or closed projects?
Archive material can be valuable precisely because it records decisions with known outcomes. The same IP, confidentiality and anonymisation checks apply.
Talk to ID43 Labs
ID43 Labs works with UK firms, starting with engineering, to capture professional reasoning, structure and anonymise it, and license it to AI labs on a revenue-share basis. You approve what is shared, data is handled under UK GDPR, and IP terms are agreed in writing. If you'd like to explore whether your firm's archive could become a new income stream, book a partnership call at id43labs.com.
See how we work with engineering firms: ID43 Labs for engineering firms.
This article is general information, not legal advice.