ClientSafe AI

Warns before possible sensitive data is sent into supported AI chat tools. Local-only scanning for professionals.

ClientSafe AI helps accountants, auditors, tax advisers, solicitors, recruiters, HR teams and other professionals reduce the risk of accidentally sending sensitive information into AI chat tools. It checks prompt text locally in your browser before you send it to supported AI sites. If it detects possible client, personal, financial, tax, payroll, HR, medical, company, password, API key or confidential information, it shows a clear warning and gives you the option to redact the prompt or go back and edit. A stricter default is used: the “send anyway once” bypass is off unless the user enables it in settings. Supported sites: ChatGPT Claude Gemini Copilot Perplexity Key features: Local-only scanning No account required No remote AI processing No analytics or tracking No external scripts Custom warning modal One-click prompt redaction Redaction success confirmation Accountant, solicitor, recruiter, HR/payroll and general privacy presets Custom client/sensitive terms Ignored terms / allowlist controls Test scanner in settings Popup site-status panel showing whether protection is active on the current tab Clear unsupported-site message Restricted content script matches; no `<all_urls>` permission Incognito disabled for this release ClientSafe AI is not a substitute for your firm’s GDPR, confidentiality, cybersecurity, professional ethics or AI-use policies. It is a practical browser safety layer to help reduce accidental disclosure risk.