Data and AI safety
Using AI shouldn't mean handing over your data. Here's what goes into AI tools, and what never does.
Putting AI to work in a business means being careful with information. These are the rules I work to on every project.
The rules
- Personal data is never entered into AI tools. Neither are credentials, keys or secrets. They're removed or masked first
- Sample or anonymised data is used to build and test wherever possible
- Your information is confidential. It's used only for your project, and I'll sign an NDA before seeing anything sensitive if you'd like
- A person checks anything uncertain or important before it's acted on
- You can see what an AI step did, and why, so it can be checked and improved
The full detail is in the Data Protection Policy.
In the systems I build for you
When a system I build uses AI to read documents or handle requests, the same care goes into its design:
- Only the information a step needs is sent to it
- Uncertain results go to a person, not straight into your records
- Access is limited to the people and systems that need it
- There's a clear record of what was processed and what happened
Questions to ask any AI supplier
Whoever you work with, it's worth asking three things: what data goes into their AI tools, whether it's used to train models, and who checks the results. Clear answers are a good sign.