Skip to content

Blog

Can ChatGPT steal your ideas? What happens to what you type, and what UK law protects

On Free, Plus and Pro plans OpenAI may train on your chats unless you opt out, and its terms say other users can get similar output. UK copyright protects how an idea is expressed, not the idea, so keep confidential details out of the prompt.

Short answer: the risk worth managing is what you choose to type. OpenAI3 may train on your content unless you opt out, and UK law gives an unprotected idea very little cover. The good news is that both of those are within your control.

Three different worries behind this question

“Can ChatGPT steal my ideas?” usually bundles three separate fears. Forum threads go round in circles because people answer different ones.

  1. Someone else sees your chat. A colleague, someone with access to your account, or OpenAI staff. That is an access question, covered in who can see my ChatGPT conversations and, for work accounts, can my employer see my ChatGPT history.
  2. Your text is used for training. Your words help shape a future model.
  3. The model gives someone else a similar idea. Another user asks a related question and gets something that looks like yours.

Each has a different answer, so take them one at a time.

What OpenAI says happens to your chats

Start with training, because it has a switch. OpenAI’s Help Center says that on its services for individuals (Free, Plus and Pro) “we may use your content to train our models”. To opt out, turn off Improve the model for everyone under Settings > Data controls, or choose Do not train on my content in OpenAI’s Privacy Portal. Either one is enough. Once you opt out, OpenAI says it won’t use your new conversations to improve its models.

There is a catch the forum threads rarely mention. If you give feedback on a response, such as a thumbs up or thumbs down, OpenAI says the entire conversation linked to that feedback may be used for training, even if you have opted out.

Temporary Chat is the second control. OpenAI says a temporary chat is not used to improve its models while it remains temporary. It stays out of your history and does not create or update memories. It is not instant deletion, though: OpenAI may keep a copy for up to 30 days for safety purposes. If you save a temporary chat, it becomes a regular chat and follows your normal training setting.

Business accounts work differently. OpenAI says that by default it does not use inputs or outputs from ChatGPT Business, ChatGPT Enterprise, ChatGPT Edu or its API to improve its models. Its Enterprise privacy page adds that business data is used for training only if the organisation has opted in to share it.

Why ChatGPT sometimes hands your idea back to you

OpenAI’s Terms of Use for UK residents say that, as between you and OpenAI, you keep your ownership rights in what you type and you own the output. The clause after that one matters more here: “Output may not be unique and other users may receive similar output from our Services.”

That is the usual explanation when a friend shows you a ChatGPT answer that sounds like your idea. Lots of people ask similar questions and get similar answers. A subscription box for dog owners, or a detective who turns out to be the narrator, will occur to thousands of people with or without a chatbot.

There is also a second, rarer mechanism. Researchers have shown that language models can memorise parts of their training data. A 2020 paper by Carlini and colleagues queried GPT-2 and extracted hundreds of verbatim text sequences from its training data, including names, phone numbers and email addresses. That model was GPT-2, trained on public web text, so the paper does not tell you how likely ChatGPT is to repeat one person’s chat. It does show why opting out of training is worth it if what you type is valuable.

What UK law protects, and what it does not

Under the Copyright, Designs and Patents Act 1988, copyright subsists in original literary, dramatic, musical and artistic works (section 1). A literary, dramatic or musical work gets no copyright “unless and until it is recorded, in writing or otherwise” (section 3). The Intellectual Property Office8’s IP Basics guidance spells out what follows: copyright protects the expression of a work, not the idea behind it. The same guidance says you cannot patent an idea or a way of thinking.

So your written chapter, pitch deck or code is protected as soon as it is recorded. The concept behind it is not. If someone else arrives at the same concept on their own, copyright only helps you if they copy your actual words, code or images.

For an unpublished business idea, the protection that fits is confidentiality. IPO guidance says the law of confidentiality protects trade secrets. To keep one protected, you must establish that the information is confidential and make sure anyone you tell signs a non-disclosure agreement. If they pass it on, that is a breach of confidence and you can take legal action. Government guidance on trade secrets sets out three conditions. The information must be secret, it must have commercial value because it is secret, and reasonable steps must have been taken to keep it secret, such as limiting sharing to those who need it. Pasting the details into a consumer chatbot with training switched on is hard to square with that, so count the prompt box as part of how carefully you keep the secret.

If your idea involves other people’s personal details, such as clients, patients or staff, data protection is a separate issue. The ICO publishes guidance on AI and data protection for organisations.

For creative work, see writing a book with AI and whether AI art is illegal. For what you can do with what ChatGPT makes, see using ChatGPT images commercially.

How to share an idea with ChatGPT more safely

Do these in order:

  1. Turn off training. Go to Settings > Data controls and switch off Improve the model for everyone, or use the Privacy Portal option. It covers new conversations.
  2. Skip the thumbs. On individual plans, feedback can put the whole conversation back into training.
  3. Use Temporary Chat for anything sensitive, and don’t save it. Remember a copy may be kept for up to 30 days for safety.
  4. Record your expression first. Write the draft, save the file and keep it dated. Copyright attaches to the recorded work.
  5. Protect the idea through confidentiality. Tell people only under an NDA, and only those who need to know.
  6. Keep the valuable specifics out of the prompt. You can get help structuring a pitch without giving the formula, the client list or the supplier terms. Describe the shape of the problem and keep the part that makes it valuable to yourself.
  7. For work ideas, use the work account. If your employer has ChatGPT Business or Enterprise, the no-training default applies there. Check your employer’s AI policy too.

If serious money or a dispute is involved, read can you use AI to get legal advice and then speak to a solicitor. This post is general information, not legal advice.

What the sources do not settle

  • How likely your idea is to turn up elsewhere. The terms admit that similar output happens, and nothing published puts a figure on it.
  • Whether a chatbot costs you confidentiality. The trade secret guidance sets out the conditions, but it does not say whether typing a secret into ChatGPT would, on its own, lose you that protection in a UK court.
  • Whether OpenAI’s policies stay the same. Both Help Center articles had been updated within the last day when we checked them in October 2026, so look again before you rely on them.

Sources

11
  1. help.openai.com1OpenAI Help Center: How your data is used to improve model performance
  2. help.openai.com2OpenAI Help Center: Temporary chat in ChatGPT
  3. openai.com3OpenAI: Enterprise privacy
  4. openai.com4OpenAI Terms of Use (EEA, Switzerland and UK)
  5. legislation.gov.uk5Copyright, Designs and Patents Act 1988, section 1
  6. legislation.gov.uk6Copyright, Designs and Patents Act 1988, section 3
  7. gov.uk7GOV.UK: How copyright protects your work
  8. gov.uk8Intellectual Property Office: IP Basics
  9. gov.uk9GOV.UK: KAM guide, intellectual property management (trade secrets)
  10. ico.org.uk10ICO: Artificial intelligence guidance
  11. arxiv.org11Carlini et al., Extracting Training Data from Large Language Models (arXiv)

FAQ

Questions people ask

Are my ideas safe with ChatGPT?

Safer than the forum threads suggest, but not risk-free. You can turn off training, and OpenAI's terms say other users may get similar output, which is different from getting your chat. The real weak point is typing confidential details you needed to keep secret.

Should I share my ideas with ChatGPT?

Share the shape, not the secret. If an idea is valuable because nobody else knows the details, keep those details out of the prompt and protect them by telling people only under a non-disclosure agreement.

Does ChatGPT train on my conversations?

On Free, Plus and Pro, OpenAI says it may use your content to train its models unless you opt out. By default it does not train on ChatGPT Business, Enterprise, Edu or API data.

Can I stop ChatGPT using my chats for training?

Yes. Turn off Improve the model for everyone under Settings > Data controls, or choose Do not train on my content in OpenAI's Privacy Portal. This covers new conversations. Giving thumbs up or down feedback can still put that whole conversation into training.

Is an idea protected by copyright in the UK?

No. The Intellectual Property Office says copyright protects the expression of a work, not the idea behind it, and the CDPA 1988 requires a literary work to be recorded. For an unpublished business idea, confidentiality and NDAs are the usual protection.

All posts · This post as markdown. General information for UK readers, not legal, financial or medical advice.