Using AI without leaking your work

Plain-English guides for anyone who suspects AI is learning their life's work. What actually happens to what you paste, and how to use AI on confidential work without handing it over.

For inventorsIs it safe to put your patent in ChatGPT?Pasting an unfiled invention into ChatGPT can count as public disclosure and put your patent at risk, with no grace period in Europe. Here is what actually happens, and how to use AI on a patent without losing it.6 min read GuideDoes ChatGPT train on what you paste?Yes, by default many AI tools can use what you paste to train their models, and the defaults change without much notice. Here is what actually happens to your data, and how to use AI without feeding it your work.5 min read For law firmsAI confidentiality for lawyers and law firmsPasting client information into a consumer AI tool can waive privilege and breach confidentiality. Here is how lawyers can use AI on sensitive matters without letting any single model see the whole file.6 min read GuideHow to stop AI from training on your workA practical guide to keeping your writing, inventions, and confidential documents out of AI training data, from opt-out settings to the stronger approach of never letting any single model see your whole work.7 min read ComparisonIPVault vs. pasting into ChatGPTA plain comparison of pasting confidential work into ChatGPT versus using IPVault, across who sees your document, training, sensitive-part control, and proof. See why splitting beats trusting one vendor.5 min read For writersStop AI from training on your screenplayPasting an unproduced script into an AI tool can feed your story to a model that learns from it. Here is how screenwriters can get AI notes and coverage without handing over the whole screenplay.5 min read For foundersUse AI on your startup's confidential plansYour business plan, cap table, and unreleased metrics are exactly what you do not want an AI to learn. Here is how founders can use AI on confidential plans without handing the whole thing to one model.5 min read GuidePrivate AI that does not train on your dataThe safest private AI is not one vendor you trust, it is a setup where no single model ever sees your whole work. Here is what to look for in a private AI for confidential documents, and why single-vendor promises fall short.5 min read GuideCan ChatGPT be used against you in court?Your AI chats can be records, and records can be discoverable. Here is how what you paste into ChatGPT could be used against you or your client, and how to keep sensitive material out of the picture.5 min read GuideWhat AI tools do with your data: the 2026 state of playA plain summary of what is publicly documented about how AI tools handle what you paste, from default training to changing terms and live lawsuits, and why the safest assumption is that anything you send could be used.6 min read GlossaryThe IP-Safe AI glossaryPlain-English definitions of the terms behind using AI without leaking your work: public disclosure, training data, zero-knowledge, masking, opt-out, on-device, privilege, and IP-Safe AI.5 min read