Chrome Extension Blocks Secret Pasting Into AI Tools

VaultBix is a Chrome extension that warns or blocks users when they paste secrets, personal data, or proprietary code into AI tools. The company says detection runs locally in the browser, with no account required and no data sent to a server.

VaultBix, a Chrome extension from two high school seniors, is built to stop users from pasting sensitive data into AI chat tools. The product says it can detect API keys, JWTs, social security numbers, credit cards, private keys, and other secrets before they leave the browser.

The company launched the tool after seeing engineers, including the founders themselves, paste production secrets into ChatGPT while debugging. On its site, VaultBix says “every paste is a one-way leak” and argues that telling people not to paste sensitive material is not a real security control.

⚡ New to this?

This is a browser extension meant to stop people from accidentally sending secrets into AI chat tools. A secret is something like an API key or login token, which can be used to access a service if it is exposed. The bigger issue is that many AI tools process pasted text on their servers, so a mistake can become a security incident.

🦞 OpenClaw angle

If your team uses self-hosted agents or browser-based AI tools, add paste-time secret detection before prompts ever reach the model. Put the check in the browser or client side first, and make the default action block or redact high-risk values like API keys, tokens, and private keys. If you run an internal tool, log only hashes or masked snippets so incident records do not become another secret store.

VaultBix is open source and runs entirely locally, according to the company. It says the extension does not send prompts to any server, does not require an account, and does not collect tracking data. The free version also makes no network requests for detection, and the company says it uses regex and entropy analysis to identify secrets in about 1 millisecond per request.

The extension currently supports ChatGPT, Claude, Cursor, Copilot, Gemini, Perplexity, and other browser-based AI tools. VaultBix says it uses 19 explicit host permissions rather than a broad all-sites permission, and that users can choose between strict, balanced, and passive modes. In strict mode, the tool blocks a request; in balanced mode, it warns on critical leaks.

VaultBix says it detects more than 45 types of secrets and personal data. The patterns listed on its site include AWS access keys, OpenAI keys, Anthropic keys, GitHub personal access tokens, Stripe live keys, GitLab tokens, Slack tokens, DB connection strings, SSH or RSA private keys, emails, phone numbers, credit cards, and social security numbers.

The company also says it keeps incident logs in hashed form rather than storing raw secret values. For teams, VaultBix says a paid tier is planned with org-wide policy controls, an incident dashboard, SSO, audit logs, custom detection rules, and SIEM integration. Those features are listed as coming soon.

The founders said the product was built after about five months of work and roughly 45 detection patterns. They also said they are adding more supported sites over time and plan to bring browser protection to IDE-based agents later.

VaultBix is available as a free Chrome extension, with the team tier and enterprise features still in development.

Source: HN Show HN ↗

More from Security News