AI Detection

The privacy risks of online AI detectors

What happens to text you paste into an AI detector, the questions to ask of any tool before you use it, and extra care for student work and confidential documents.

By aitextdetectly team3 min read

Pasting text into an online AI detector sends it to someone else’s server. Often that’s fine. But the text people check is often exactly the kind that shouldn’t travel far: a student’s essay, an unpublished manuscript, a client report, a job application. Before you paste, it’s worth knowing where the words go.

We make one of these tools. Read our answers below with the same questions you’d put to anyone else.

Five questions to ask of any detector

Most of the answers are in the tool’s privacy policy or terms. If they aren’t, that’s an answer too.

1. Is the text stored, and for how long? Some tools keep every scan. That can be a feature, since a scan history is handy, but a stored copy can be breached, subpoenaed, or kept longer than you expect. Look for a stated retention period, not just “we take security seriously”.

2. Is it used to train models? Your text might be used to improve the detector or other products. Check whether training is on by default, whether you can turn it off, and whether free and paid plans have different terms.

3. Who else receives it? Many detectors send text on to another company that runs the underlying model. Privacy policies often list these as “subprocessors” or “service providers”. Each one is another place your text goes.

4. Does it go into a shared database? Plagiarism checkers work by comparing a document against earlier ones, and some can keep submitted documents to compare future submissions against. If a tool combines AI detection with plagiarism checking, find out whether your document is added to that collection and whether you can opt out.

5. Can you delete it? Look for a way to delete your scans and your account, and what deletion actually removes.

Student work needs extra care

Teachers are among the heaviest users of AI detectors, and student writing comes with its own rules. In the US, student education records are protected under FERPA, and many schools and districts keep a list of approved tools that staff may upload student work to. Before pasting a student’s essay into any detector, check your school’s policy on approved tools. The student’s work may also include personal details, such as names, health information, or family circumstances, that the tool doesn’t need.

Confidential and personal documents

Contracts, client deliverables, and internal reports may be covered by a confidentiality agreement that doesn’t allow sharing them with outside services. Job applications and personal statements often hold details about someone’s life that they’d share with a hiring manager but not a software company.

Ways to reduce the risk

  • Remove names, contact details, and other identifying information before you scan. The detector doesn’t need them to judge the writing.
  • Prefer tools that state they don’t keep your text and don’t train on it, and that name who else processes it.
  • Use a tool your school or employer has approved when you’re checking someone else’s work.
  • Be careful with scanning only an excerpt to share less. Short passages are harder for any detector to judge, so the result becomes less reliable.

What we do with your text

When you scan, your text goes to our detection provider, TypeSafe, to run that one request. We don’t keep a copy of the text or the result, we don’t sell it, and we don’t use it to train models. Text you type on the homepage before signing in stays in your browser’s session storage until it’s handed to the detector. You can delete your account, credits, API keys, and workspace data at any time from the Account page. The privacy policy has the full list, including the other services we use for sign-in email and payments.