Privacy policy

Effective date: October 26, 2026
Last updated: November 1, 2026

Who we are

Sight4, LLC ("Sight4", "we", "us") makes the Sight4 browser extension and the Sight4 desktop helper for AI coding tools. You can reach us at info@sight4.com.

The short version

Sight4 checks what you're about to send to an AI tool on your own device.
Your prompts, pastes and files are never sent to Sight4. We don't have a server that receives them.
Sight4 doesn't decrypt or intercept your internet traffic.
Today there's no Sight4 account, and there are no analytics and no tracking in the extension, the desktop helper or this website.
What Sight4 keeps (settings, and short records of the kind of data it found) stays on your device, and you can delete it.
Anything you choose to send to an AI site or tool still goes to that company, just as it would without Sight4. Their privacy policy covers it, not ours.
We're building a console where an account owner can see and manage the devices they add. It will get device names, versions, which AI tools are installed, usage counts once reporting is on, and what kind of data was caught, never what you wrote. See what it will hold.

For a one-table view of who can see what, see Your data, plainly.

The browser extension

What it does

The extension checks prompts on ChatGPT (chatgpt.com), Claude (claude.ai), Gemini (gemini.google.com), Grok (grok.com, and Grok inside X at x.com/i/grok), Kimi (www.kimi.com and kimi.moonshot.cn) and Muse (muse.ai). When you send, paste or drop text into the prompt box, it looks for things like API keys, passwords, payment card numbers, bank account numbers (IBAN), US bank routing numbers, US Social Security and taxpayer (ITIN) numbers, Canadian Social Insurance Numbers, Medicare numbers (MBI), US health provider numbers (NPI and DEA), and, on the Strict setting, email addresses and US and Canadian phone numbers. If it finds one, it stops the send or asks you to check first. When a number you paste only might be a password, a US passport number or a bank account number, it blurs it and asks you about it before it's sent. It never blocks it. If you turn on "Check attached files" (off by default), it also checks files you attach on those sites.

Checking Grok inside X is optional and off until you turn on "Also check Grok inside X", which asks your browser for access to x.com. When it's on, the extension is allowed to run on the whole of x.com, because X loads Grok without a full page change. It checks nothing and records nothing outside Grok. Muse is checked only in test builds; the version in browser stores doesn't run on muse.ai.

How your data is handled

Prompts and pasted text are read inside the extension on your device, checked, and then dropped. In this version they're never stored, logged or sent anywhere by Sight4. My activity, once it ships and only if you turn it on, keeps your own flagged prompts on your device, encrypted (see below).
Attached files (only with file checks on) are read on your device for one check: the first few bytes, a ZIP file's list of file names, the text of a text file up to 4 MB, or a small scaled-down copy of an image. File names are used only to spot medical-image words and file types. Nothing from a file is stored or sent.
Sight4's notices never show the data they found, only what kind it was (for example "Payment card number").

What it stores, and for how long

Everything below is kept in your browser's own extension storage on this device. None of it is uploaded.

Your settings: the protection level (Balanced or Strict), whether file checks are on, and whether blocking AI sites is on. Kept until you change them or remove the extension.
This session's counts: how many times Sight4 warned or blocked, per site and per kind of data, and whether each open tab is protected. Kept in memory only and cleared when you quit the browser.
Your choices on "possible" matches: when Sight4 blurs something that might be a password, a passport number or a bank account number and you choose to keep or remove it, it records the day (not the time), the site, the kind of data and your choice. Never the value. Kept for 7 days, then deleted automatically. The extension's menu shows these records and has a Clear button.

What it never does

In this version, it makes no network connections of its own.
In this version it doesn't download code or rules. The detection rules are built in.
It has no analytics, ads or tracking, and doesn't read your browsing history.
Its code runs only on the AI sites listed above (on x.com, only if you turn that on, and then it loads site-wide but acts only in Grok), and on Google Search if you turn on hiding Google's AI answers below.

Hiding Google's AI answers (optional)

The extension has a switch, off by default, that hides AI Overviews and links into AI Mode on Google Search. When you turn it on, your browser asks you to give Sight4 access to www.google.com. The extension saves whether the switch is on, in your browser only. When you turn it off, that access is removed, unless Blocking AI sites still uses it for Google's AI Mode (below). Access you give for blocking doesn't turn this switch on. On Google, Sight4 reads the results page in your browser only to find those AI panels and hide them. It doesn't record your searches or send them anywhere. Google can tell that an extension hid something on its page.

Blocking AI sites (optional)

The extension has a "Block AI sites" setting, off by default. You can block every AI site on Sight4's built-in list, including sites added in later updates, or choose individual sites. Your choice is saved in your browser only. Your browser itself blocks the sites you chose and shows its own "blocked" page. The browser does the matching, so Sight4 doesn't see or record which sites you visit. If you choose Google's AI Mode, it isn't blocked but sent to Google's plain web results instead, which is why choosing it asks for access to www.google.com. If you don't give that access, AI Mode is blocked like the other sites. Giving it doesn't turn on hiding Google's AI answers.

What the AI site can still see

The AI site's own page can read what you type in its prompt box, even before you press send and even when Sight4 stops the send. It can also tell that a Sight4 notice appeared and what you clicked, but not what the notice says. Sight4 can't prevent this, and the AI company's own privacy policy applies to it.

The desktop helper for AI coding tools

The Sight4 desktop helper (a command-line program called aisec) runs on your computer as you, without administrator rights (an organization install needs them once, to set it up). It plugs into the hook settings of Claude Code, Cursor and Codex, so those tools ask it to check each prompt before sending, check which MCP servers and tools the AI wants to use, and check what those tools send back.

No traffic interception. It doesn't decrypt, proxy or inspect your internet traffic. It only sees what the AI tool hands it through the tool's own hook feature.
Nothing is uploaded. In this version the helper makes no network connections.
Your prompts aren't written down. In this version, prompt text, the values it finds and your custom keywords are never written to its logs or output.
A short local log. For each decision it writes a line to a log file on your computer: the tool, the action taken, the protection level, the kinds of data found and how many, the MCP server and tool name (with anything that looks like a secret or your custom keywords replaced), a fixed reason code, the rule version, and the time. The file is readable only by your user account and is capped at about 1 MB, with one older file kept (on Windows, it can grow past that if another program holds the file open).
Settings and backups. Your policy file and a backup of the few hook settings it changes in Claude Code, Cursor or Codex are kept in the same folder. It copies only the hook settings it changes, never your env or apiKeyHelper entries. If one of your existing hook commands contains a secret, the backup holds that line too.
Short-lived session notes. To spot an AI agent hammering one website, it keeps a small file per AI session: which website domains the agent's commands reached and how often, and whether the session saw signs of planted instructions or read a credential file. Never full web addresses, commands, or the session's own ID. A file unused for 24 hours is cleared the next time a session starts. The folder also holds a random key, made on your device, that keeps those file names from revealing the session.
Install records. When it's installed, it keeps a record of what it installed and a short install log, so it can be cleanly removed.

The log and settings live in ~/Library/Application Support/Sight4/aisec/ on macOS, %LOCALAPPDATA%\Sight4\aisec\ on Windows and ~/.config/Sight4/aisec/ on Linux. Deleting that folder deletes everything the helper keeps. When an organization installs it for everyone on a computer, the install record, backups and the organization's policy file go in a system folder only administrators can change (/Library/Application Support/Sight4/aisec/ on macOS, C:\ProgramData\Sight4\aisec\ on Windows). For those installs, removing everything also means deleting that folder, which needs administrator rights.

The helper also has an inventory command that lists AI tools, MCP servers and browser extensions on your computer, the programs connected to AI services, and five security settings (like disk encryption and firewall). You can run it yourself to see the result. Once the helper is installed, it also runs this inventory when you log in and every hour, and saves the latest result in the same folder, along with fingerprints (hashes) of AI skill files so it can tell when one appears or changes. It doesn't send any of it anywhere.

Claude Code and Cursor keep their own history of your prompts. That's controlled by those tools, not by Sight4.

This website and the waitlist

This site sets no cookies and uses no analytics, trackers, ads, third-party fonts or third-party scripts.
Do Not Track and Global Privacy Control. We don't track you across websites, so there's nothing for these browser signals to turn off. If we ever add something they apply to, we'll honor them.
The site is hosted on Microsoft Azure, which may keep standard server logs (such as IP address, browser type and the page requested) to run and protect the service.
Waitlist form. When you join with the form on our home page, we keep three things: your email address (in lowercase), which option you picked (home and family, organization or small business, or developer), and the date you joined. The form itself doesn't store your IP address, browser details or anything else from the request, and it has no free-text box. Our web host's own server logs, described above, are separate. Our storage service also records the time each entry was written; we don't use it.
Where it's kept. The list is kept in Sight4's own Microsoft Azure storage account in the United States, not with a mailing-list or sign-up service. Only Sight4 staff can read it. The form adds you to the list and can't read it back.
We don't confirm addresses. We don't send a confirmation email, so someone could add an address that isn't theirs. If you get an email from us about a waitlist you didn't join, reply or write to info@sight4.com and we'll remove it.
How we use it. Only to tell you about Sight4 early access. We send those emails through our email provider, Postmark, which gets your address for that purpose only. We don't sell it or share it for advertising, and we'll delete it if you ask.
Leaving the list. Every waitlist email has a one-click unsubscribe link. When you use it, Postmark stops sending to you and your entry is deleted from our list automatically. Postmark keeps your address on its do-not-send list so you aren't emailed again by mistake.
Joining by email. If you'd rather email info@sight4.com, we get your email address and whatever you write, and use it the same way.

The console and My activity (not built yet)

We're building a console where an account owner can see which of the devices they've added are covered and set the same protection for each. None of this exists yet. Before it launches, we'll update this page with the exact list of what's sent, and the app will show it to you too.

What a console will hold

What devices send: device names; app, rule and policy versions; which AI tools and Sight4 apps are installed; and detection details: what kind of data was found, what Sight4 did, which AI tool or MCP tool was involved, and when. Never prompts, pastes, files, commands or the data that was found. Once a device is connected to a console, its admins see these.
Usage counts (how much AI tools are used) only once usage reporting is turned on. It's off by default, and each device shows a notice before counts are shared.
Connection details: the console also sees connection details such as IP addresses, kept in short-lived logs to run and protect the service.
Alerts: admins get alerts about security events, such as a policy change no admin approved. Alerts never include what anyone wrote.
Where it runs: we host your console.

Who controls console data

For your Sight4 account and billing, Sight4 decides how the data is used. For console data, the account owner decides. We handle that data only to run the console for them. If you have questions about console data, ask the person who added your device first; we'll help them answer.

Your account and billing

Every plan, including Free, will need an account. We'll keep your name and email address (see Creating an account), plan, licenses and a random ID for each device, and for paid plans billing details from our payment processor, to run your account and check licenses. We won't hold your full card number; the payment processor does.

Policy signing

We sign each policy change with a separate key for your account, kept in Microsoft Azure Key Vault, and your admins can see a log of every signature. On Organization, you can keep the key in your own cloud instead.

My activity

My activity is an optional view, off by default, that lets you look back at your own flagged prompts, with the caught data masked. It's kept on your device, encrypted, and never sent to Sight4 or to the account owner who added your device.

Accounts aren't open yet. When they launch, this is how signing in will work.

Who runs sign-in. Microsoft Entra External ID, part of Microsoft Azure, handles sign-in for us. If you sign in with a code sent by email, Microsoft sends that email. It keeps sign-in logs, including your IP address, for about 7 days to run and protect the service. It also keeps its own audit logs, which record things like a sign-in record being created or deleted. These logs stay with Microsoft for their set time, even after we delete your sign-in record.
Signing in with Google, Apple or Microsoft. If you choose one of these, that company learns that you sign in to Sight4. From Google we receive only your name, email address and Google account ID (the "openid", "email" and "profile" permissions). We use them only to run your Sight4 account. We never sell them or use them for advertising, and we'll delete them if you ask. Sight4's use of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements.
Text message codes (planned). We plan to offer codes by text message later. If you choose them, Microsoft keeps your phone number with your sign-in record, and Microsoft and the mobile carriers that deliver the message see your phone number, its country and when a code is sent. Any extra sign-in step that Microsoft runs, such as an email or text message code, is held by Microsoft, not Sight4.
Extra checks for important changes (planned). We plan to let you confirm important changes with a passkey or an authenticator app. When we add them, we'll keep only the public key of a passkey (your device keeps the private key), an authenticator app's shared secret, encrypted, and only a hash of each recovery code, not the codes themselves. These aren't built yet, so Sight4 keeps none of them today.

Microsoft makes a sign-in record first. When you start signing up, External ID creates a user record for you in Sight4's sign-in directory. It holds your email address, your name and, if you sign up with Google, Apple or Microsoft, the ID of that account. This happens before we ask your age or where you live.
While you answer our questions. The console keeps your name and email address for up to 30 minutes, only to fill in the sign-up form for you. They're encrypted with a key that only the sign-up cookie in your browser can unlock, so they can't be read from our database. The console also keeps the random ID Microsoft gave your sign-in record and which sign-in method you used. Every 5 minutes, the console deletes sign-ups that have run out.
If you answer No to "Are you 18 or older?", or you live outside the US and Canada. No account is created, and the console keeps nothing about you except, until the deletion is finished, the random ID Microsoft gave your sign-in record. We ask Microsoft to delete that record right away. About 5 minutes later, we ask Microsoft to remove it for good from its deleted items. If that last step doesn't go through, Microsoft removes it by itself within 30 days. If a deletion keeps failing, we keep trying, and then someone on our team finishes it by hand.
If you stop partway. If you reach our questions and leave without answering, your sign-up runs out after 30 minutes, and within about 5 minutes after that we ask Microsoft to delete your sign-in record, the same way as above. We don't if you've created an account by then or started signing up again; if a sign-up of yours is still open, we wait for it. If you leave on Microsoft's own sign-up pages, before reaching our questions, your sign-in record isn't deleted automatically yet; we're building a daily cleanup for these.
If you create an account. We keep the name and email address you confirm, where you live (United States or Canada), your language, when you said you're 18 or older (never your birth date), which version of the Terms you accepted and when, and the random ID of your sign-in record and the sign-in method you used. On your account record, we keep only the day (not the time) you last signed in and last used Sight4.
Staying signed in. Your browser keeps a session cookie. We keep only a hash of it, never the cookie itself, along with when you signed in and when the session was last used. A session ends when you sign out, after 7 days without use, or after 30 days at most, and we delete it then.

We don't sell your personal information, and we don't share it for advertising. We use a few service providers to run Sight4, and they may handle data only for us:

Microsoft: Azure hosts this website and, when it launches, the hosted console. Azure Table Storage holds the waitlist. Microsoft Entra External ID will handle sign-in when accounts launch, and holds a sign-in record for each account holder, and for anyone who starts signing up until it is deleted (see Creating an account).
Postmark delivers our email: waitlist and early-access emails now. When accounts launch, it will also deliver codes Sight4 sends itself, such as one to confirm your contact email. Sign-in codes sent by email come from Microsoft.
Stripe will process payments when paid plans open. Stripe holds your card details; we don't.

How long we keep it

On your device: until you clear it or remove Sight4. The extension's password choices are deleted after 7 days.
Waitlist: until you ask us to delete your entry, or until early access is open to everyone and we no longer need the list, whichever comes first. Either way, we delete it no later than 24 months after our last early-access email. The same goes for waitlist emails you send us.
Email we send through Postmark: Postmark keeps a history of each message (who it went to, the subject, and whether it was delivered or bounced) for 45 days, even if we delete your waitlist entry sooner. If you unsubscribe, or an email to you bounces or is marked as spam, Postmark keeps your address on its do-not-send list with no end date, so you aren't emailed again, until you ask us to remove it.
Website server logs: for as long as our web host keeps them to run and protect the service.
Sign-ups that don't finish (when accounts launch): the console deletes your sign-up details within about 5 minutes of the sign-up running out (30 minutes after it starts). Microsoft's sign-in record is deleted as described in Creating an account. Microsoft's sign-in logs are kept for about 7 days.
Console connection logs (when the console launches at app.sight4.com): up to 30 days, to run and protect the service.
Account and billing data (when accounts launch): while your account is open, and after that only as long as tax and accounting rules require.
Hosted console data (when it launches): for the period your plan sets, up to 12 months, then deleted. It's deleted when you close your account.

Children

Sight4 accounts are for adults 18 and older, and the service isn't directed to children. You need to be 18 or older to join the waitlist and, when accounts open, to create an account.

We don't knowingly collect personal information from anyone under 18. If you answer No to "Are you 18 or older?" when signing up, no account is created and we ask Microsoft to delete the sign-in record it made (see Creating an account). If we learn that an account holder is under 18, we'll email them to say so, then delete the account and its data within 30 days, except records the law requires us to keep. If we learn we have other personal information from someone under 18, such as a waitlist entry, we'll delete it. If you think someone under 18 has joined the waitlist or written to us, email info@sight4.com and we'll delete it. We won't ask you to prove your relationship to them.

An account owner may add devices that other people use, including people under 18. If you add a device someone else uses, you're responsible for having the right to do that and for any notice or consent the law requires. In this version, the extension and the desktop helper don't send personal information about anyone to Sight4, and a device's prompts are never sent to Sight4 or to the account owner.

Your choices and rights

Extension data: use Clear in the extension's menu, or remove the extension, which deletes everything it stored.
Desktop helper data: delete the folder listed above. Running aisec hooks uninstall removes its hook entries from Claude Code and Cursor.
Waitlist: email info@sight4.com from the address you signed up with to see, correct or delete your entry, or anything you emailed us. We'll answer within 30 days.
Leave the list: use the one-click unsubscribe link in any waitlist email. It deletes your entry from our list automatically, and Postmark stops sending to you. You can also reply to any of them, or write to info@sight4.com.

Depending on where you live, you may have more rights over your personal information, such as the right to access, correct, delete or move it, or to object to how it's used. For Sight4 today, those requests cover your waitlist entry, the emails you send us and any server logs our web host keeps. Once accounts launch, they also cover your account and billing data. For console data, the account owner who added your device decides, and we'll help them answer.

We answer these requests wherever you live, not only where a law requires it. We may ask you to confirm a request from the email address it's about. If we turn a request down, we'll tell you why, and you can reply to ask us to look at it again. If you still disagree, you can contact your state attorney general or, in the EU or UK, your data protection authority.

We don't sell your personal information, and we don't share it for cross-context behavioral advertising. If you live in California, you have the right to know what we collect, to delete or correct it, and to opt out of its sale or sharing. Since we do neither, there's nothing to opt out of. We won't treat you differently for using any of these rights.

If you're in the EU or UK

Sight4, LLC is the controller for your waitlist entry and the emails you send us, and, when accounts launch, for your account and billing data. We use your information only where we have a reason the law accepts: to answer you and keep you posted when you join the waitlist (our legitimate interest, or your consent), to run your account and subscription (our contract with you), and to keep tax and accounting records (a legal duty). For hosted console data, we act on behalf of the account owner. Sight4 is based in the United States, so your information may be handled there; where the law requires it, we use safeguards such as standard contractual clauses. You can also complain to your local data protection authority.

Security

The safest data is data we never receive, so Sight4 is built to keep your prompts on your device. In this version, the extension makes no network connections of its own, and the helper's log file is readable only by your user account. No system is perfect: if you find a security problem, please tell us at info@sight4.com.

Changes to this policy

If we change what Sight4 collects, we'll update this page and the effective date before the change ships. For a significant change, we'll also say so in the extension's release notes and email the waitlist.

Contact us

Sight4, LLC · info@sight4.com