Privacy Policy - Emet

Last updated: 20 August 2026 · Applies to version 2.2.3 and later

Emet is a browser extension that transcribes the audio of the tab you are watching, identifies factual claims in the Hebrew speech, and displays an assessment with links to sources. We do not create an account for you, do not identify you, and do not store your audio or transcripts.

Two modes

• Shared service (default). So that you don't need API keys of your own, requests are relayed through a Cloudflare Worker we operate at emet-proxy.emet.workers.dev, which holds our provider keys and forwards to Groq (Whisper transcription and LLM), Google Gemini, and Serper (Google search results, region IL). In this mode your audio and transcripts do pass through infrastructure we control.
• Your own key (Advanced settings). If you select a model and enter your own API key, requests go directly from your browser to that provider. Our server is not involved and sees nothing.

Stored on your device only

• Your API keys, if you supply any, in chrome.storage.local. Never sent to us, and never to anyone other than the provider each key belongs to.
• Preferences: chosen model, panel position and size, hidden cards.
• A daily usage-minutes counter, used to enforce the free quota. A number only - no history of what you watched.
• Session state (active tab, content title) in chrome.storage.session, cleared when you stop or close the browser.

What is sent, and when

Nothing is sent before you click "Start fact-checking". From then until you stop: tab audio chunks go to the transcription service; transcript segments and the page title go to the language model for claim detection and verification; search queries go to the evidence-search service. In the default mode these are relayed via our Worker; in bring-your-own-key mode they go directly to the provider you chose.

What our server sees and keeps

• Audio, transcripts and verdicts are never written to disk. Each request is processed and forgotten. We have no way to reconstruct what you watched or what was said.
• There is no user identifier. Requests carry no name, email, account or device ID, and we cannot link one person's requests together.
• Your IP address is used for rate limiting only. Cloudflare supplies it, and the server uses it to increment a daily counter so a single user cannot exhaust the shared budget. That counter expires automatically after 48 hours. It is not linked to any content and is not used for identification, profiling or advertising.
• No cookies, pixels, analytics or tracking - in the extension or on the server.

What we do not do

We do not sell or share data - we have none to sell. We do not use your data for any purpose other than operating the extension. We do not use it to determine creditworthiness or for lending purposes. We do not collect browsing history: the extension runs only on the page where you explicitly started it. We do not train models on your content.

Third-party services

Google AI / Gemini · Groq · Serper · Deepgram · Anthropic · Bright Data · Cloudflare

Permissions

tabCapture - capture tab audio, only after you click Start. offscreen - process audio in the background; MV3 service workers cannot access the Web Audio API. storage - save your keys and preferences locally. activeTab - inject the results panel into the page you are watching and read its title to identify speakers.

Children

The extension is not directed at children under 13 and does not knowingly collect information from them.

Changes

If we change how data is handled, we will update this document and the date above, and notify you inside the extension before the change takes effect.

Compliance statement

Our use of information received from Chrome APIs, and our transfer of that information to any other party, adheres to the Chrome Web Store User Data Policy, including the Limited Use requirements.

Contact

For questions about this policy, open an issue on the project's GitHub page.