Thai PDPA Guard – Mask personal data before AI chat

Thai employees paste customer data into ChatGPT, Claude and Gemini every day — national ID numbers, names, phone numbers, addresses, bank accounts — and the organization never knows. Thai PDPA Guard is built specifically for Thai personal data: it detects and masks it before the message leaves the text box. ■ What it does • Scans the chat input of ChatGPT, Claude and Gemini as you type • Thai 13-digit national ID / tax ID — validated with the official mod-11 checksum, not just digit counting • Thai mobile and landline numbers, emails, credit cards (Luhn), Thai bank accounts, license plates, passports • Thai addresses (tambon / amphoe / province / postcode) and Thai names with titles • Masks with meaningful placeholders such as [บุคคล_1] [บัตรปชช_1] [เบอร์_1] — the AI still understands the context • Automatic unmask: when the reply contains [บุคคล_1], you see the real name on screen • 3 modes: warn (default) / auto-mask / block send until handled (for organizations) • Uses on-device Gemini Nano when Chrome supports it to catch names without titles — works fully without it ■ Privacy • Everything runs on your device. The codebase contains no network calls at all. • The placeholder↔value table lives only in the tab's session memory and is wiped when the tab or browser closes. • No account, no analytics, no storage of what you type. Only counters such as "12 items masked". ■ Why Thailand's PDPA carries administrative fines up to THB 5,000,000, and a real case has already resulted in THB 7,000,000 in combined fines. This extension is a technical safeguard a DPO can roll out today without changing how staff work.