Locked learning resources

Join us and get access to thousands of tutorials and a community of expert Pythonistas.

Unlock This Lesson

Locked learning resources

This lesson is for members only. Join us and get access to thousands of tutorials and a community of expert Pythonistas.

Unlock This Lesson

What Gets Sent, and Who Sees It

Everything the app puts in the context gets sent to the company that runs the model. That’s a lot more than what you type. The context includes files you upload, files and emails the agent reads, pages it visits, and any saved memories the app adds.

Where It Goes

The models behind popular assistants run on the company’s computers. Each time the app sends the context, it travels over the internet to those computers, where the model reads it and writes a reply.

That’s true even when the agent works on your own computer, as with the client contract in Where Is Your Agent Working? Your files stay where they are, but whatever the agent reads goes into the context, and the whole context gets sent to the model. If an agent opens ten documents to answer one question, the text of all ten goes along.

What Happens to It There

Locked learning resources

Join us and get access to thousands of tutorials and a community of expert Pythonistas.

Unlock This Lesson

Already a member? Sign-In

Locked learning resources

The full lesson is for members only. Join us and get access to thousands of tutorials and a community of expert Pythonistas.

Unlock This Lesson

Already a member? Sign-In

00:00 What Gets Sent, and Who Sees It. Everything the app puts in the context gets sent to the company that runs the model. That’s a lot more than what you type.

00:10 It includes files you upload, and files and emails the agent reads. It also includes pages it visits, and any saved memories the app adds.

00:20 The models behind popular assistants run on the company’s computers. Each time the app sends the context, it travels over the internet to those computers.

00:29 There, the model reads it and writes a reply.

00:33 That’s true even when the agent works on your own computer. Your files stay where they are. But whatever the agent reads goes into the context, and the whole context gets sent to the model.

00:45 If an agent opens ten documents to answer one question, the text of all ten goes along. What happens to it there depends on the product, your plan, and your settings.

00:56 That covers whether the company keeps your conversations. It covers how long they’re kept, and whether they’re used to train future models. Work and school accounts often follow different rules from personal ones.

01:08 Many apps have a setting for training. It’s often in a privacy or data section of the settings. And it doesn’t always use the word training, so look closely.

01:18 People may see it, too. Depending on the company’s policy, its staff may review some conversations, for example to check for misuse. A conversation you share by link can be read by anyone who has the link.

01:31 And on a work account, your employer may be able to see what you’ve sent. For anything an app’s settings don’t show, like how long conversations are kept, the company’s privacy page is the place to check.

01:43 A few habits keep sensitive information out of the context. Keep passwords, ID numbers, and card numbers out of the chat. The agent doesn’t need them, and once they’re sent, you can’t take them back.

01:55 Think before you give an agent a whole folder or inbox. It may read more than you had in mind, and whatever it reads gets sent. And take out details the task doesn’t need.

02:05 For example, before you share a client list, remove the parts the agent doesn’t need.

02:11 So whatever the agent reads gets sent, and the settings decide part of what happens next. Below this video, you can practice in a sample app. Find the training setting and turn it off.

02:22 Remove an old shared link. Then ask the assistant whether your chats still get sent.

Become a Member to join the conversation.