Category
Fundamentals
The key concepts to understand privacy in the age of AI.
- 6 min readCan You Put Your Resume in ChatGPT?A resume is a full identity sheet. What leaves for the AI when you paste it, the two cases (your resume / a candidate's), and how to anonymize it before ChatGPT.
- 6 min readCan You Put Your Medical Records in ChatGPT?Pasting your test results or medical report into ChatGPT is possible, but these are health data. How to understand a term without exposing your identity.
- 6 min readCan You Paste a Work Email into ChatGPT?A work email carries several people's identities. What you expose by pasting it into ChatGPT, and how to anonymize it before the prompt.
- 6 min readCan You Put a Bank Statement in ChatGPT?Pasting a bank statement into an AI is possible, but it reveals your whole financial portrait. What it really exposes, and how to anonymize it before the prompt.
- 6 min readCan You Put a Screenshot in ChatGPT?A multimodal AI reads your whole screenshot, not just the part you meant. The real risk of third-party data in the background, and the right reflex.
- 7 min readWhich AI Tools Train on Your Data by Default? (Comparison)It depends on the tool AND the plan. Consumer tiers often train by default; API and enterprise rarely do. An attributed, dated comparison.
- 7 min readReversible Anonymization: Use AI While Reducing Your Data's ExposureReplace each sensitive value with a reversible token, keep the mapping on your device, restore everything in the answer. ONYRI's differentiator, explained honestly.
- 6 min readCan You Put a Bank Details Slip (RIB) in ChatGPT?A RIB alone can't empty your account — that's a myth. The real risk of an exposed IBAN, and how to anonymize it before ChatGPT.
- 6 min readCan You Put a Payslip in an AI Tool?A payslip concentrates name, address, salary and often the social security number. Pasting it raw into an AI exposes it all at once. The simple fix.
- 6 min readCan You Put a French Kbis Extract in an AI Tool?A Kbis is a public record anyone can order from the registry. The real AI risk isn't the document itself — it's the director's name and the context you paste around it.
- 6 min readCan You Put Your French Tax Notice in ChatGPT?AI can explain your French tax notice, but don't paste the raw document. The tax number and notice reference act as an authentication key. The fix: anonymize before the prompt.
- 7 min readIs Your Data Safe With American AI Providers?Most leading AI is American. Handing it personal data is a transfer outside the EU. What the GDPR says, where the Data Privacy Framework stands, and the fix that keeps your data under control.
- 6 min readCan AI Clone Your Voice? What You Need to KnowA few seconds of your voice, pulled from a video or a voicemail, can be enough to make a convincing clone. What the FTC and the GDPR say, and the simple defenses that work.
- 6 min readDoes AI Record Your Voice? (Voice Mode, Assistants, Voiceprints)When you talk to a voice mode or an assistant, your audio leaves your device. What happens to it, what the law says, and how to keep sensitive data off the mic.
- 6 min readAI Privacy for Small Business: A Practical PlaybookSmall businesses move fast and paste customer, staff and financial data into consumer AI with no policy or review. What the GDPR, ICO and CNIL say, and a concrete 5-step playbook.
- 7 min readAre Local AI Models More Private Than the Cloud?A model running on your own machine never sends your prompt out: a real privacy win for the inference step. But local has limits. The balanced answer, and the middle ground.
- 7 min readIs It Safe to Use AI for Legal Advice in the UK? (SRA)AI hallucinates law, and English courts have sanctioned lawyers over fake AI citations. A consumer chatbot gives you no privilege either. What the SRA expects — and the fix.
- 7 min readHere Are the 7 Jobs Most at Risk of an AI Data LeakHealthcare, law, finance, HR, developers, support, marketing. Seven jobs ranked by data sensitivity — and the common fix that protects them all.
- 9 min readTop 10 Types of Sensitive Data You Should Never Put Into AIThe ten kinds of data to keep out of ChatGPT, Claude and Gemini — ranked by risk, each with a concrete example, plus the one fix that covers the whole list.
- 7 min readTop 5 AI Privacy Myths, DebunkedA VPN, paying, deleting the chat, temporary mode, “nothing to hide”: five reassuring beliefs, all false. The ranking, the facts, and the one fix that holds.
- 7 min readCan AI Guess Your Personal Information From What You Type?A model doesn't need your name to know who you are. It reads your style, your words, your small details — and infers your personal data. What the research shows, and the only fix.
- 7 min readWhat Is Data Minimization (and Why It Matters for AI)?Data minimization is one of the seven GDPR principles: collect and process only what is strictly necessary. What it means, why it lowers risk, and how to apply it to an AI prompt.
- 7 min readCan Anonymized Data Be Re-Identified?Erasing the name isn't enough. A few quasi-identifiers are enough to find a person in an “anonymized” dataset — landmark demonstrations, what the GDPR says, and what actually protects you.
- 10 min readAI Privacy: The Complete Guide to Protecting Your DataWhat AI does with your data, the real risks, the protections that hold and those that don't — and the one lever you truly keep: what you send.
- 6 min readIs It Safe to Ask ChatGPT Medical Questions?Asking ChatGPT about your health stacks two risks: reliability (it can be wrong) and confidentiality (ultra-sensitive data, outside HIPAA, retained). What actually protects you.
- 7 min readDoes AI Share Your Data With Third Parties?"Sell" no, "share" yes. Who AI providers pass your data to — vendors, affiliates, courts, acquisitions — and the sharing you trigger yourself.
- 7 min readIs It Safe to Use AI for Legal Advice?Two dangers hide behind “asking ChatGPT for legal advice”: an AI that invents law, and the total absence of attorney-client privilege. What really happens, and the only fix.
- 6 min readWhat Personal Data Does ChatGPT Collect About You?Account, prompts, files, device, IP, Memory profile, third-party sources: the full inventory of what OpenAI collects per its privacy policy — and the one lever that stays in your hands.
- 7 min readIs ChatGPT Safe for Therapy? What Happens to What You ShareOpenAI's CEO says it himself: there's no legal confidentiality when you use ChatGPT as a therapist. Why your mental-health exchanges aren't protected — and the only fix that holds.
- 5 min readAI now models the brain: why data minimization mattersAI no longer just handles text: it models brain activity. What Meta's TRIBE v2 shows — and why the simplest reflex, hand over only what's needed, has never mattered more.
- 6 min read“I have nothing to hide” — the wrong lens for AI privacyThe “nothing to hide” argument collapses the moment you look at what you actually paste into AI: rarely your secrets, often other people's. Privacy isn't a confession.
- 6 min readCan an AI repeat what you typed?“It gets lost in the crowd” is a myth. Three peer-reviewed studies measure how AI models memorize and emit data — and why you shouldn't paste it in the clear.
- 6 min readThe hidden cost of a single leaked promptPasting one sensitive value feels costless. The bill is hidden below the waterline — and the brain is wired not to see it.
- 6 min readWhy careful people still paste their data into AIEveryone knows they shouldn't. And everyone does it anyway. It isn't a knowledge problem, it's a friction problem — and that changes the solution.
- 6 min readPrompt injection: how your data can leakAn instruction hidden in a document or web page can hijack an AI and make it reveal what it holds in context. Understand prompt injection and shrink the exposed surface.
- 6 min readSensitive data: United States vs France, what changesNational identifiers don't look alike from one country to the next. A US vs France overview — and why good detection must be country-aware.
- 6 min readWhy in-browser processing protects your dataThe best-protected data is the data that never leaves. How in-browser (client-side) processing cuts the risk surface at the source.
- 6 min readAnonymization vs pseudonymization vs tokenizationThree confused terms, three different legal statuses. The clear guide to anonymization, pseudonymization and tokenization — and what it changes for AI.
- 7 min readWhat is sensitive data? A practical taxonomySensitive data isn't just names and emails. Here is a 6-family taxonomy — from names to API keys — so you know exactly what to protect before sending it to an AI.