Tracehold Free 作成者: Tracehold

Tracehold is a privacy guardian for when you use artificial intelligence.

How many times have you pasted text into ChatGPT without checking what was in it? A password, an API key, a customer email, a card number… Once sent, there is no way back.

Tracehold Free analyzes what you write BEFORE you press Enter and, if it detects sensitive data, blocks the send and warns you. You get one last chance to remove that data before sharing it with an AI.

Analysis happens in your browser and your content never leaves your device. Tracehold Free has no account, needs no sign-up and does not send the content of your messages to any server. Text analysis runs 100% in your browser. The detection ruleset ships inside the extension package (it is not downloaded over the network). The only connections to app.tracehold.com are anonymous: a statistics ping (at most once every 12 hours), small batches of anonymous scan events, an optional "Is this a false positive?" report you send explicitly, a service-configuration lookup (URLs only, no identifiers) and a single anonymous uninstall notice. None of them include the text of your messages. Your data is never used to train models.

WHAT IT DETECTS
• Passwords and secrets: variables like PASSWORD=, SECRET=, API_KEY=, .env files.
• Vendor keys: AWS, Google, GitHub, OpenAI, Stripe, Slack and similar.
• Tokens and certificates: JWT, private keys (-----BEGIN PRIVATE KEY-----).
• Connection strings with user and password (protocol://user:password@host).
• Personal data: credit cards (validated), IBAN, national IDs, emails and
phone numbers.

WHERE IT WORKS
Tracehold Free activates on the supported AI assistants — 46 provider
families, including ChatGPT, Claude, Gemini, Microsoft Copilot, GitHub
Copilot, Perplexity, Mistral, Grok, DeepSeek, Poe, Meta AI, Notion AI,
HuggingFace Chat, Grammarly, QuillBot and many more — and recognizes a
curated catalog of ~2,460 AI domains. The detector loads on any page but
unloads itself instantly except on recognized AI assistants; network access
stays limited to the providers listed in the manifest (no <all_urls>).

HOW IT WORKS
1. Install Tracehold Free and open your favorite AI.
2. Write or paste your message as usual.
3. If there is sensitive data, Tracehold blocks the send and shows it to you.
You decide: edit the message or send it anyway.

No configuration, no buttons to press. It works from the first minute.