Most of us click "Allow" on a camera prompt without reading it. The call connects, someone's face appears, and we stop thinking about it.
That's fine most of the time. But a lot of apps now describe themselves as private, secure or encrypted, and those words get stretched until they barely mean anything. A family call, a doctor's appointment over video, a chat with someone you met online, a job interview: each one carries a slightly different level of risk, and the label on the app tells you surprisingly little about which risks it actually covers.
This guide is for people who don't work in tech and don't want to. No jargon you'll need to look up. Just what is really going on when you turn your camera on, what the common privacy claims mean, where they stop being true, and how to check any video platform yourself in about ten minutes.
You won't need to become paranoid. You'll just know what to look for.
What your browser actually hands over when you click "Allow"
When a website asks to use your camera, the request comes from your browser, not the website itself. That's a good thing. The site can't quietly switch on your camera. It has to ask, and the browser decides whether to show the prompt.
When you say yes, here is what you are granting, in plain terms: that website, on that address, gets to receive a live feed from your camera (and usually your microphone, if you allowed both) while you have it open.
A few details are worth knowing.
The permission is tied to the site's address. If you allow your camera on one video service, a different website doesn't automatically get it. Each site has to ask separately.
The permission often sticks around. Many browsers remember your answer, so next time you visit, the camera turns on without a prompt. Some browsers offer a one-time option, often labelled something like "Allow this time" or "Allow while visiting the site". If you see it, that's usually the better choice for a service you don't use often.
The site can't run your camera in secret. Modern browsers show an indicator whenever the camera is live, usually a small icon in the address bar or on the tab. Many laptops also have a hardware light next to the lens. If that light is on and you didn't expect it, close the tab.
What the permission does not give: access to your files, your photos, your other tabs, or your contacts. A camera permission is a camera permission. The worry about "the app seeing everything" usually comes from other permissions, granted somewhere else, often to a phone app rather than a website.
Phone apps are a different story. An installed app asks the operating system for permission, and on both iPhone and Android you can see and change every permission each app holds. Background camera use is tightly restricted on recent versions of both systems, but it's still worth checking which apps have camera access at all. You may be surprised.
How to take the permission back
Revoking camera access takes less than a minute. The exact menu names shift a little between versions, but these paths will get you close.
Chrome on a computer: click the small icon at the left end of the address bar while you're on the site, then switch the camera off. To see every site at once, go to Settings, then Privacy and security, then Site settings, then Camera.
Safari on a Mac: open Safari's Settings, choose the Websites tab, then Camera in the left column. Each site you've visited is listed with its current setting.
Firefox: open Settings, go to Privacy & Security, scroll down to Permissions, and click Settings next to Camera. You can remove individual sites from the list.
Your computer itself: on a Mac, System Settings, then Privacy & Security, then Camera shows which apps (including browsers) can reach the camera at all. On Windows 11, it's Settings, then Privacy & security, then Camera.
iPhone: Settings, then Privacy & Security, then Camera lists every app with access. Camera permissions for websites in Safari live in Safari's own section of Settings.
Android: look for a Permission manager under Settings, usually inside Security and privacy or Privacy. Manufacturers label it slightly differently, but it lets you browse by permission type and see every app holding camera access.
A useful habit: once every few months, open that list and remove anything you don't recognise or no longer use. It takes two minutes and closes doors you forgot you opened.
Is video chat encrypted? Usually, yes. The real question is which kind
Here's the good news. If you're using a reasonably modern video service in a browser, your call is almost certainly encrypted while it travels across the internet. Most browser video calling is built on a technology called WebRTC, and WebRTC doesn't allow unencrypted audio or video. Encryption isn't an optional extra a developer can forget to switch on. It's part of how the thing works.
So "is video chat encrypted?" usually gets a yes. The trouble is that "encrypted" can describe at least three different situations, and they protect you from very different things.
Encrypted in transit. Your video is scrambled as it moves between your device and wherever it's going next. Someone snooping on the coffee shop Wi-Fi, or anywhere along the cables, sees noise. This is the baseline, and it's what most services mean.
Encrypted at rest. Data stored on a company's servers (your messages, your profile, maybe recordings) is scrambled on disk. This protects against someone stealing a hard drive. It doesn't stop the company itself from reading the data, because the company holds the keys.
End-to-end encrypted. Only the people on the call hold the keys. The video is scrambled on your device and unscrambled only on the other person's device. Any server in the middle passes along gibberish it can't read, even if it wanted to.
Most of the confusion here comes from services that offer the first, mention the word "encryption" in their marketing, and let you assume the third.
Encrypted in transit vs end-to-end: the courier and the code
Think of it this way.
Encryption in transit is like sending a letter through a courier who carries it in a locked case. Nobody on the street can read it. But the case gets opened at the courier's depot, where the letter is sorted, and then it's locked into a new case for the second leg of the trip. The depot staff could read your letter if they wanted to. You're trusting them not to.
End-to-end encryption is like writing your letter in a code only the recipient can crack, and then handing it to the courier. The depot still handles it. They just can't read a word.
On a video call, the "depot" is the company's server. Whether your video passes through one depends on how the call is set up.
In a peer-to-peer call, your device and the other person's device connect directly, or as directly as the network allows. The encryption is negotiated between the two devices, so no company server is unlocking your video along the way. Sometimes a direct connection isn't possible, because of strict office firewalls or certain mobile networks, and the call gets routed through a relay server instead. In standard WebRTC setups, that relay forwards the encrypted packets without unlocking them. It's a mail slot, not a depot.
There's a catch worth understanding, even if you never act on it. In most consumer apps, the two devices find each other and swap their security details through the company's server. You're trusting that server to introduce you honestly. Apps that offer full end-to-end encryption usually add a way to check this, like a security code both people can compare. Most peer-to-peer calls don't offer that check, which is why careful services describe them as encrypted and peer-to-peer rather than calling them end-to-end.
In a server-routed call, your video goes to the company's media server, which then sends it on to everyone else. With ordinary transport encryption, that server unlocks the video, handles it, and locks it again. This is how most group calling works, and it's the main reason "encrypted" and "end-to-end encrypted" are not the same promise.
Neither setup is automatically bad. Server-routed calls are often more reliable and can offer features that peer-to-peer calls can't. You just want to know which one you're using, so you know who you're trusting.
Why group calls break end-to-end encryption in ways one-to-one calls don't
If end-to-end encryption is better, why doesn't every app use it for everything? Group calls are a big part of the answer.
With two people, the maths is simple. Your device sends one video stream to one other device. Even a phone on a weak connection can handle that.
Now add people. If five people are on a call and everyone sends their video directly to everyone else, each device is uploading four copies of its video and downloading four streams at the same time. At eight or ten people, ordinary home internet and most phones start to buckle. Video freezes. Batteries drain. Phones get hot.
So group calls almost always use a central media server. Each person sends one stream up, and the server copies it out to everyone else. That's efficient, and it's why large meetings work at all. But it means the server is, by default, sitting in the middle of every stream.
To make a group call truly end-to-end encrypted, the app has to add a second layer of encryption on top, one the server can't remove. That's possible, and some major apps do offer it, but it creates real headaches:
Keys have to be shared among everyone. Every participant needs the means to decrypt everyone else's video, and that has to be handed out securely to each person.
People join and leave. When someone drops off, the keys should ideally change, so a person who left can't keep listening if they somehow still receive the stream. Doing that smoothly in a live call is hard engineering.
Features break. Cloud recording, live captions, processing done on the server, dial-in phone numbers: many of these need the server to see the actual video or hear the actual audio. Switch on end-to-end encryption and they tend to be switched off or limited.
Trust gets harder to check. In a big group, verifying that everyone is who they claim to be, and that no silent extra participant has been added, gets harder with every person.
This is why it's common to see apps where end-to-end encryption exists but is off by default for group meetings, or where turning it on disables features you rely on. That's not necessarily bad faith. It's a genuine trade-off.
One-to-one calls avoid most of this. With two people, there's no technical need for a server to sit in the middle of the media, there's only one other person to share keys with, and nobody is joining or leaving halfway through. A one-to-one call set up as peer-to-peer is already much closer to the privacy people imagine when they hear the word "encrypted", even if it doesn't carry the full end-to-end label.
If privacy matters for a particular conversation, a one-to-one call is usually the safer format, for purely technical reasons.
What a platform can and can't see
Even when your video is well protected, the company running the service knows a lot about the call. Encryption hides the content. It doesn't hide the fact that the conversation happened.
Here's what a typical platform can see, regardless of how the video itself is protected:
Who you talked to. Account names, user IDs, sometimes phone numbers or email addresses.
When, and for how long. Start time, end time, how often you call the same person.
Roughly where from. Your IP address, which reveals an approximate location and your internet provider.
What device. Browser, operating system, phone model, sometimes screen size and language settings.
Text you type. In-call chat and direct messages are often handled separately from the video, and they're frequently not end-to-end encrypted even when a call is. Check this one specifically.
Payments. If the service charges money, it knows what you paid, when, and how. The payment company knows too.
Anything you upload. Profile photos, files shared in chat, your bio.
And here's what it can't see, if the call is genuinely peer-to-peer or properly end-to-end encrypted: the actual pictures and sound of your conversation.
That split is worth sitting with for a moment. For a lot of people, the metadata is the more sensitive part. A record showing you spoke to a particular person for an hour every Tuesday night can say more about your life than the conversation itself.
Good platforms collect only the metadata they need to run the service and keep it for a limited time. Weaker ones collect everything because storage is cheap and it might be useful one day. The privacy policy is where you find out which kind you're dealing with, and we'll get to how to read one quickly.
Screen recording detection: what it can catch and what it can't
Some video services advertise that they detect or block screen recording. This sounds reassuring, and it can be a real feature, but it has hard limits that the marketing rarely mentions.
Start with apps on phones. Both major mobile systems give app developers some tools here. On iPhone, an app can check whether the screen is currently being captured and react, for example by blanking the video. On Android, an app can mark its screen as secure, which blocks normal screenshots and screen recordings of that app and shows a black area instead. Those protections hold up reasonably well against the phone's built-in recording tools.
Browsers are a different matter. A website running in a browser has much less control. It generally can't see what other programs on your computer are doing, so it can't reliably tell whether a separate screen recording app is running. What a website can do is notice certain signals inside the browser itself: the page losing focus, some kinds of capture started from within the browser, or behaviour that commonly comes with recording tools. A well-built service can catch common browser-based capture attempts and respond by warning, blurring or ending the call.
What nothing can catch:
A second device. If someone points another phone at their screen, no software on earth can detect it. People call this the "analog hole", and there's no fix for it.
Capture hardware. External devices that grab the video signal between a computer and its monitor are invisible to the software on that computer.
Determined people with technical skill. Modified browsers, virtual machines and similar tricks get around most detection.
So the honest version of this feature is: it makes casual recording harder and catches the common methods. It doesn't make recording impossible. Any service that claims it can stop all recording is overstating what the technology can do.
The practical lesson is simple. Never show anything on a video call that you couldn't live with being saved, whatever the app promises. Detection is a speed bump, not a wall.
What a platform keeps after you hang up
The call ends. The window closes. What's left?
The video itself: usually nothing. Unless the call was recorded, the video stream isn't saved anywhere by default on most services. In a peer-to-peer call, it never passed through the company's server to be saved in the first place. If a service does record calls, for instance meeting tools with cloud recording, that should be obvious in the app and spelled out in the policy.
The call record: almost always. Who called whom, when, and for how long. Services keep this for billing, fraud prevention, troubleshooting and handling safety reports. Some keep it for months, some for years.
Text messages: usually. Chat history tends to stick around unless you delete it, and deleting it on your side doesn't always delete it from the company's servers.
Payment records: for a long time. If money changed hands, tax and accounting rules in many countries require businesses to keep transaction records for several years. Even if you delete your account, a record of your payments typically stays. That's normal and lawful, but it's worth knowing that "delete my account" rarely means every trace disappears.
Server logs. Technical logs with IP addresses and timestamps, kept for security. Well-run services clear these out after a set period.
Safety and moderation records. If you reported someone, or someone reported you, those reports and any evidence attached are often kept longer, for obvious reasons.
If you live in the EU or the UK, data protection law gives you the right to ask a company what it holds about you and, in many cases, to have it deleted. Several US states, including California, give residents similar rights. A good service makes this easy, with a clear email address or a button in your account settings. A service that makes it hard is telling you something.
How to read a privacy policy for the three things that matter
Nobody reads privacy policies front to back, and you don't need to. Open the policy, use your browser's find function (Ctrl+F on Windows, Cmd+F on a Mac, or "Find in page" on a phone), and look for three things.
1. Are calls recorded, and who can see the content?
Search for: record, recording, content, monitor, review.
You're looking for a clear statement about whether calls are recorded and whether staff or automated systems ever review what's said or shown. Some services review content for safety reasons, and they should say so plainly. Watch out for vague lines like "we may process communications to improve our services", which can cover almost anything.
Green flag: a direct statement such as "we do not record calls", or a specific description of when and why content gets reviewed.
Red flag: no mention of recording at all, on a service whose entire product is calls.
2. Who gets your data, and is it sold?
Search for: share, third parties, partners, sell, advertising.
Every service shares some data with suppliers, like hosting companies and payment processors. That's unavoidable. What you want to know is whether your data goes to advertisers or data brokers, and whether the company says it sells personal information.
Green flag: a short list of supplier types, each with a clear purpose, and a plain statement that personal data is not sold.
Red flag: "trusted partners" with no explanation, or talk of "marketing partners" and "personalised advertising" on a service where you'd be sharing your face.
3. How long is it kept, and how do you delete it?
Search for: retain, retention, delete, deletion.
A good policy gives actual time periods, or at least clear rules ("we delete call logs after 90 days", "we keep payment records for as long as the law requires"). It also tells you exactly how to request deletion.
Green flag: specific periods and a clear deletion process.
Red flag: "we retain information for as long as necessary" and nothing else. That sentence means "as long as we like".
Those three searches take about five minutes and tell you more than reading the whole thing carefully. While you're there, check the date at the top. A policy that hasn't been touched in years, on a service that has obviously changed in that time, suggests nobody is paying attention.
The platforms worth trusting are specific about their limits
Here's something that sounds backwards but holds up well in practice: the services worth trusting are often the ones that tell you what they can't do.
Security marketing is full of absolute words. Unbreakable. Fully secure. Military-grade. One hundred percent private. Real engineers almost never talk this way, because they know every system has edges. When a company describes its protection in precise, limited terms, it usually means someone who understands the technology wrote the description, and that person was allowed to be honest.
So when you read a service's privacy or security page, look for specifics:
Does it say which kind of encryption it uses, in transit or end-to-end, and does it avoid implying the second when it means the first?
Does it explain when calls go peer-to-peer and when they're routed through a server?
Does it describe what its recording protection can actually catch, instead of claiming to stop all recording?
Does it say what it keeps and for how long?
Some services do this well. For one-to-one calls, Privé Moi is a useful example: it presents itself as a privacy-first video chat platform and describes its calls as using encrypted WebRTC transport, peer-to-peer where possible, while saying it detects common browser-based screen capture attempts rather than promising to stop every kind of recording. That's a narrower claim than "fully encrypted and unrecordable", and it's more useful precisely because it's narrower. You know what you're getting, and you know where the edges are.
Compare that to an app that calls itself "100% secure" without explaining a single technical detail. It may be perfectly fine. But you've been given nothing to check, and that should make you more careful, not less.
A practical checklist before you sign up for any video platform
Run through this before you create an account, especially if you'll be talking to people you don't already know. Most items take a minute or less.
Before you sign up
Find the privacy policy and do the three searches above: recording, sharing, retention. If you can't find a privacy policy at all, stop there.
Check what the service says about encryption. Look for the words "in transit", "end-to-end" or "peer-to-peer". If it only says "secure" or "encrypted" with no detail, assume it means in transit only.
Check whether one-to-one calls are handled differently from group calls. If privacy is your priority, one-to-one is the stronger format.
See who runs the service. A real company name, a country of registration and a way to contact them. Anonymous operators are a bigger risk, whatever their policy says.
Look at what sign-up asks for. Does a private video chat service really need your phone contacts, your exact location, or your photo library? The less it asks for, the less there is to lose.
When you first use it
Choose "allow this time" for the camera if your browser offers it.
Watch the address bar for the camera indicator, and your laptop's camera light. Confirm both go off when you leave the call.
Check the in-app privacy settings. Who can see your profile, who can contact you, whether you appear in search, whether your online status is visible.
Find the block and report buttons before you need them. If you can't find them, that tells you how much attention the service pays to safety.
On an ongoing basis
Review camera and microphone permissions every few months on each device you use.
Delete conversations and accounts you no longer use. Request deletion through the service if the button in the app doesn't seem to do the whole job.
Use a separate email address for services you're trying out. If one leaks or starts sending spam, your main inbox stays clean.
If a service fails items one, two or four, move on. There are plenty of options, and the ones that take privacy seriously aren't shy about explaining how they do it.
A few habits that matter more than any setting
The technology can only protect you so far. A handful of simple habits close the gaps no app can.
Mind your background. The most common privacy leak on video calls has nothing to do with encryption. It's the letter on the fridge, the street sign visible through the window, the school logo on a jumper hanging off a chair. Take a quick look at what's behind you before you join.
Treat every call as potentially saved. Encryption protects you from strangers in the middle. It does nothing about the person on the other end. If you wouldn't be comfortable with something being kept, don't say it or show it, whatever the app promises about detection.
Keep your software updated. Browser and phone updates regularly fix security problems in exactly the parts video calls rely on. Turning on automatic updates is probably the single most effective privacy step most people can take.
Be careful on shared computers. If you use a library computer or a friend's laptop, sign out afterwards, and if you granted camera access to a site, remove it before you walk away.
Cover the lens when you're not using it. A small sliding camera cover costs almost nothing and makes the whole question of what's on or off irrelevant. Plenty of laptops now come with one built in.
The short version
Privacy on a video call isn't one feature you either have or don't. It's a few separate questions: what your browser has allowed, how the video is encrypted and who holds the keys, what the company records around the call, what it keeps afterwards, and how honest it is about all of it.
Most services encrypt calls in transit. Fewer offer end-to-end encryption, and group calls make that genuinely hard. One-to-one calls, especially peer-to-peer ones, keep the company out of the middle of your video far more easily. No service can stop someone with a second phone, so the last line of defence is always your own judgement.
You don't need to be an engineer to check any of this. Ten minutes with a privacy policy, a look at the camera permissions on your devices, and a healthy suspicion of words like "unbreakable" will put you ahead of most people.
And the next time an app asks to use your camera, you'll know exactly what you're agreeing to.