AI Policy

1. Where AI is used

AI is used in four places in TrainAR. When a job is recorded, computer vision and speech recognition read what the camera sees and what the engineer says. When a session is built, models turn those recordings into an ordered set of steps. When someone follows a session on site, models decide which step to show, and when. And when someone asks a question out loud, a model answers it using the material in their organisation's own library.

AI is not used to score, rank, rate or discipline individual workers, and it is not used to make decisions about a person's employment, pay or progression.

2. What it learns from

Our models are trained and improved using material generated on the platform, including recordings, transcripts, annotations and the corrections people make while using a session. This is how the guidance gets better over time and how the platform learns the difference between a job done well and a job done badly.

We do this under the agreement signed with each customer organisation, which sets out the permission granted to us. Where an organisation has not granted that permission, we do not use its material for that purpose.

We do not train models on material captured while no session is running. The App does not capture audio or video when it is closed or when no session is active.

3. What never leaves your business

Your recordings stay in your own tenant of the platform. We do not disclose your raw recordings, or anything that identifies your business, your customers, your staff, your premises or your commercial arrangements, to any other customer.

What improves across the platform is the model's general understanding of the work, not your material. A session belonging to your business is never shown to another business, and no output made available to another customer identifies you.

4. A person is in charge

TrainAR guides, it does not authorise. The engineer on site decides whether to follow a step, and can stop, skip or ignore any guidance at any time. Sessions are built from the work of your own senior engineers, and your organisation decides which sessions are published to its team and who can see them.

Where a session is used as evidence of competency, sign-off is made by a person, not by the platform.

5. Safety-critical work

TrainAR is used on work that carries real risk, including gas, electrical and work at height. Nothing in the platform replaces the qualifications, registrations, regulations, manufacturer instructions or professional judgement that apply to that work.

Guidance shown by TrainAR is training support. It is not a substitute for being competent and qualified to carry out the task, and it does not certify that a job has been carried out correctly or safely. If guidance conflicts with a regulation, a manufacturer's instruction or your own procedure, those take precedence and the guidance should be reported to us.

6. Accuracy and limits

AI systems make mistakes. A model can misread what the camera sees, miss a detail, or answer a question confidently and be wrong. We test our models before release and monitor them in use, but we do not represent that guidance is complete, accurate or suitable for every situation.

If you find guidance that is wrong or unsafe, report it to team@trainar.ai. We review reports of unsafe guidance as a priority.

7. Third-party AI providers

Some AI processing is carried out by third-party providers on our behalf. They act on our instructions, are bound by data processing agreements, and are not permitted to use your material to train their own models for their own purposes. A current list of our subprocessors is available on request at team@trainar.ai.

8. Security and access

Material is encrypted in transit and at rest. Each organisation's data is logically isolated from every other organisation's. Access by our staff is limited to those who need it for support, engineering or safety review, and is logged.

9. People in recordings

Recordings contain people at work. How that personal data is handled, and the rights that apply to it, are set out in our Privacy Policy. Where we film at a customer's site for marketing or case study purposes, that is separate from the platform and requires the written consent of each individual who appears.

10. Changes to this policy

Technology and our products change, and this policy will change with them. Material changes will be posted on this page and, where a change materially affects a customer organisation, notified to that organisation.

11. Contact

Questions, concerns or reports about our use of AI: team@trainar.ai, or TrainAR Limited, 22 Siskin Grove, Leighton Buzzard, LU7 4DQ, United Kingdom.