Is it safe to use ChatGPT at work?

I use AI at work every day. Claude, mostly. I am dyslexic, and it is the thing that finally got what is in my head onto the page.

So I am not going to tell you to stop.

But people ask me this a lot. Is it safe? Will it leak my data? Can I get into trouble? Should I tell my boss?

It is safe when you do four things. Here they are.

1. Find out what your workplace allows

Do this first. Before the email goes, not after.

Your organisation may well have rules on AI. They are often buried somewhere in the IT pages. Find them and answer three questions.

Which tools am I allowed to use? What information must never go into them? Do I have to say when I have used AI?

Find out who approves new tools, too. If nobody can tell you, ask your manager and write down what they said. Then you have an answer you can point to.

2. Keep private information out

This is the “will it leak my data” question.

The UK’s National Cyber Security Centre explains it plainly. What you type is visible to the company that runs the tool. It is stored, and it will almost certainly be used to develop their service at some point.

So treat the box you type into like an email to a stranger.

These stay out:

• client names and details
• anything about a colleague
• health, pay or performance
• passwords and log-ins
• anything marked confidential
• figures your company has not published

If it is about a person, data protection law already says how you must handle it. Putting it into an AI tool does not switch that off.

The fix is often simple. Take the name out and write “Client A”. Take the figure out and write “X”. Most of the time the task still works.

And check which version you are on. The free app on your phone and the version your company pays for may come with different terms. Your IT team will know which one you are meant to use.

3. Check what comes back

This is where people actually get caught out.

AI can be wrong and sound completely sure of itself. I know this better than most. This summer one of the AI agents I work with told me it had written thirty four checks to prove a piece of work. It had written none. I had already approved the work.

So check before you use anything. Every number. Every name. Every date. Any rule or law it quotes. Check them against two other sources, not against the AI.

If you cannot check it, do not send it.

And some jobs should not go to AI at all. A decision about a person is the obvious one. Who gets hired, who gets a warning, whose complaint is upheld. A tool can produce a result that is unfair to a whole group of people without anybody meaning it to. That is a job for you.

4. Say when you have used it

Should you tell your boss you use ChatGPT? If your workplace has a rule, follow it. If it does not, I would say yes.

Using a tool well is part of doing the job. Hiding the tool is what makes it look like cheating.

Your name is on the work. When it goes out, it is yours, whatever helped you write it.

Can you get fired for using ChatGPT at work? I cannot answer that for your employer. Read your policy. In my experience the risky part is rarely the tool. It is what went into it, and what came out unchecked.

When it goes wrong anyway

You pasted something you should not have. Or a wrong answer went out under your name.

Tell the person responsible. Quickly, and in whatever way your workplace asks. Write down what happened and what they said back.

I like mistakes, the bigger the better. A big one you can see, and you can fix it. The dangerous ones are the quiet ones nobody reports, which sit there for a year.

Then change one thing about how you work, so it does not happen again.

Where we come in

Future Fluency is the qualification we are writing for exactly this. It is still a draft. At Level 3 there is a whole unit on working safely and responsibly with AI, and it is built on your own job.

You find your own workplace rules. You use AI on your real tasks, inside them. You keep a record of what you did. You spot a risk in your own work and report it. Then you look back at where you fell short and what you changed.

If you want the quick version of how to ask AI for good work, I wrote a tip sheet on talking to AI at work.

Your first Future Fluency module is free. After that, learning is paid.

→ Start your free first module

Sources

• UK National Cyber Security Centre, ChatGPT and large language models: what’s the risk? (queries are visible to the provider, stored, and likely to be used to develop the service): ncsc.gov.uk
• Information Commissioner’s Office, guidance on AI and data protection: ico.org.uk