Skip to content
The Coach's Lab
The Scouting Report: Issue #10

The setting you never actually chose

September 13, 2026 5 minutes read
AI ToolsPrivacySystems

A choice you probably never made

There is a setting in every one of these tools that decides whether your conversations get used to help build future versions of the model. Claude put that decision in front of people directly at one point. ChatGPT just has it switched on unless you go find it and turn it off. Either way, most coaches I talk to have no idea where it sits on their own account.

I don't think that is careless. Nobody has time to read the fine print on every piece of software they use. But this particular setting has consequences worth two minutes of your attention, especially if you work with athletes.

What actually happens to what you type

On the consumer plans, Claude, ChatGPT, and Gemini can all use your conversations to help improve their models unless you turn that off. Business, team, and enterprise accounts work differently and are not used for training at all, which is a real distinction if your school or organization pays for the tool.

On Claude, that same setting also affects how long your conversations are kept. Leaving training on means Anthropic can hold that data for five years. Turning it off shortens that considerably, though the specific number they used to publish is not in the current policy anymore. ChatGPT works differently on this point, because turning training off does not change retention there at all. Your conversations stay on their servers indefinitely until you go delete them yourself.

There is one more thing worth knowing, and it is the reason I would rather you handle this now than later. Once your data has gone into a training run, it cannot be pulled back out. Opting out stops what happens going forward. It does not undo anything that already happened, which means you should turn this setting off ASAP.

The part that matters more if you coach at a school

Everything above applies to you personally. There is a second layer if you work for a high school, a university, or anywhere else that has rules about student records.

Athlete information at a school is usually protected in ways that your own notes are not, and your institution almost certainly has a policy about where that information can and cannot go. Consumer AI accounts are not built for regulated data, and that is not a knock on the tools, it is just what they are. If you are pasting anything that identifies an athlete into a personal account, you are making a decision on behalf of your institution that you probably were not asked to make.

I am not going to pretend to know your school's policy, because they vary and I am not the person to interpret it for you. What I would do is ask. Your compliance office, your athletic administrator, or whoever handles data agreements will know whether there is an approved tool, and if there is, that is the one to use for anything athlete-related.

Inside the Lab

Two things to do, the first takes under a minute.

Go change the setting. In Claude it is under Settings, then Privacy, where you will find a toggle for helping improve the models. In ChatGPT it is Settings, then Data Controls, then the option about improving the model for everyone. In Gemini, look for the activity setting. Turn them off unless you have a reason to leave them on.

While you are there, a couple of other things are worth knowing. Both tools have a mode for conversations you would rather not keep, called incognito in Claude and temporary chat in both ChatGPT and Gemini. These chats don't get used for training even if your main setting is switched on. That is a good option for a one-off you do not want sitting in your history. Since ChatGPT holds your conversations until you delete them, going through and clearing out old ones is worth doing.

Then get in the habit of separating your work. The rule from a few issues back still does most of the heavy lifting here. Describe the shape of what you are working with rather than uploading the thing itself. Use athlete IDs or initials instead of names. Keep anything involving actual athlete records out of a personal account entirely, and use whatever your institution has approved for that.

Everything else, your programming, your notes, your learning, your presentations, is yours and it is fine.

Where the coach stays in charge

I want to be clear that none of this is a reason to stop using these tools. I use them every day and this newsletter exists because I think they are worth the time. The point is that you should know what you agreed to, because right now most coaches do not.

The thing I keep coming back to is that the risk here is almost entirely about athlete information, and that piece is completely avoidable. You can get every benefit these tools offer without ever putting a kid's name into one. That is not a compromise or a workaround, it is just how the work should be set up anyway.

The Coach's Scouting Report

  • Signal. Once data goes into a training run it cannot be removed, so this is worth handling today rather than eventually. On Claude, leaving the toggle on also means your conversations can be held for five years.
  • Noise. Assuming a paid plan means your data is private. Assuming your personal account is an appropriate place for athlete records.
  • Next Rep. Open your privacy settings and look at where the training toggle sits right now. Change it if you want to. Then find out whether your school or organization has an approved tool for anything involving athlete information.

See you next Sunday.

Colin

The Coach's Lab | Less busywork. More coaching.