PasteGuard — Stop Pasting Secrets into ChatGPT & Other AI Chats

On-device DLP for AI chats. Warns before API keys, SSNs, and customer data reach ChatGPT, Claude, Gemini & 5 more.

PasteGuard watches for sensitive data in text you paste into AI chat sites and warns you before it is sent. All scanning runs entirely inside your browser. No clipboard content is ever uploaded, logged, or transmitted anywhere. HOW IT WORKS When you paste text into a supported AI chat site, PasteGuard scans it instantly. If it finds anything sensitive - an API key, a social security number, a credit card number, or a confidential business term - it shows a warning panel before the message is sent. You then choose to send anyway, remove the sensitive parts, or cancel. WHAT IT DETECTS Secrets and credentials AWS access keys and secret keys, Stripe live keys and webhook secrets, GitHub tokens and fine grained personal access tokens, Slack tokens, JSON Web Tokens, PEM private keys, npm tokens, database connection strings, OpenAI keys, Anthropic keys, Google AI keys, Google service account files, Hugging Face tokens, Replicate keys, Groq keys, Perplexity keys, OpenRouter keys. Personal identifiers Social security numbers, credit card numbers (verified with Luhn checksum), IBAN bank account numbers, US passports, US employer identification numbers, US individual taxpayer identification numbers, US DEA registration numbers. International PII across 16 countries United Kingdom: national insurance numbers, NHS numbers, UTR numbers, company registration numbers, VAT numbers, PAYE references, bank account numbers. Italy: codice fiscale, partita IVA. Spain: DNI, NIE, CIF, social security numbers. Germany: Steuer-ID, VAT numbers, social insurance numbers, Handelsregisternummer. France: NIR, SIREN, SIRET, RPPS, ADELI. Netherlands: BSN, KVK, VAT numbers. Belgium: national register numbers, VAT numbers, company numbers. Portugal: NIF, NISS, cartão de cidadão. Sweden: personnummer, organisationsnummer. Switzerland: AHV numbers, UID. Brazil: CPF, CNPJ, PIS, CNH, RENAVAM, título de eleitor. Canada: social insurance numbers, business numbers. Australia: TFN, ABN, ACN, Medicare numbers. India: Aadhaar, PAN, GSTIN, UAN, voter ID, DIN. Mexico: CURP, RFC, NSS, CLABE, INE. Semantic signals (requires on device AI support in Chrome) Customer and company names, internal project codenames, confidential business content. INDUSTRY COMPLIANCE PRESETS One click presets turn on the detectors most relevant to your regulatory context. Available presets: HIPAA, PCI DSS, SOC2, GDPR, Financial Services, Legal. Applying a preset only turns on the relevant categories. It does not turn off anything you have already enabled. FILE SCANNING (PRO) Drop a text file, PDF, or Word document onto the Scan page to check it for sensitive content before uploading it to an AI tool. Findings are highlighted directly inside the document view. REGION DEFAULTS Choose your country in settings and PasteGuard turns on the detectors most relevant to your jurisdiction by default. You can enable or disable any detector individually at any time. PRIVACY PasteGuard does not make any network requests of its own. The only permission it uses is storage, to save your settings locally. No clipboard content, no paste text, and no findings are ever sent to any server. The audit log (a count of findings by severity and site) is stored locally in your browser and can be exported or deleted at any time from the options page. The full privacy policy is available at: pasteguard.com/privacy v0.2.0 (Update) PasteGuard now covers 12 AI chat platforms. International PII detection across 16 countries: United Kingdom, Italy, Spain, Germany, France, Netherlands, Belgium, Portugal, Sweden, Switzerland, Brazil, Canada, Australia, India, Mexico and the United States. Over 70 detector patterns total. Industry compliance presets: one click configurations for HIPAA, PCI DSS, SOC2, GDPR, Financial Services and Legal. Each preset enables the detectors most relevant to that regulatory context and optionally elevates their severity. File scanning: drop a text file, PDF or Word document onto the new Scan page to check it for sensitive content before uploading it to an AI tool. PDF findings are highlighted directly in the document view. On device AI layer: if your device supports Chrome's built in language model, PasteGuard now runs a secondary semantic scan to catch customer names, internal codenames and confidential business signals that pattern matching alone cannot detect. Detection rules redesigned: detectors are now grouped by category with filter chips and a search bar so you can find and toggle specific rules quickly. Region picker: the country selector is now a compact flag and name grid instead of a long list. Send button interception: PasteGuard now also intercepts the send button and the Enter key, not just the paste event, so content typed or dragged in is also checked before it is sent.