These words get used as if everyone already agrees what they mean. Below are our working definitions — technically accurate, without jargon for its own sake. Every entry has its own link, so our articles can point straight at a definition instead of explaining the same term twice.
Jump to: Age verification · Algorithmic feed · ALPR · Client-side scanning · Content moderation · Data broker · Deplatforming · DSA · End-to-end encryption (E2EE) · Facial recognition · Geoblocking · Metadata · Shadowban · Threat model · Tor · Transparency report
Age verification
Any system that checks whether a user is old enough to see something — from a self-declared birth date, through an AI estimate of a selfie, to a scan of a government ID. The strong methods identify you; the private methods only estimate, and estimates fail hardest at the edges. The industry prefers the term “age assurance”, which blurs that difference.
Why it matters: a law written to protect children can quietly turn an anonymous web into an identified one.
Read more: California Steps Back From Dangerous Expansion of its Age-Gating Law
Algorithmic feed
A timeline ordered by a ranking system that predicts what will hold your attention, rather than by time. Its inputs are your behaviour, the behaviour of people it considers similar to you, and the platform’s own commercial goals. The weights are not published and change constantly.
Why it matters: nothing has to be deleted for you to never see it.
Read more: Hate “The Algorithm?” RSS Is One of the Tools You’ve Been Looking For
ALPR — automated licence plate reader
A camera and software that read every passing plate and record it with a time and a location, whether or not the vehicle is suspected of anything. The records are usually pooled into searchable databases shared between agencies and private operators, and kept for months or years.
Why it matters: put enough of them on enough poles and you have a searchable history of where a car — and by extension a person — has been.
Read more: Are Your Local Police Using Flock Safety ALPRs to Scan for Immigrants?
Client-side scanning
Checking content on your own device, before it is encrypted and sent, against a list of prohibited material — normally by comparing digital fingerprints rather than by reading the file. Because the check happens before encryption, it works on end-to-end encrypted messengers without technically breaking the encryption.
Why it matters: it moves the search from the network into your pocket, and somebody else decides what the list contains.
Content moderation
How a platform decides what stays up: automated classifiers at scale, human reviewers for the difficult cases, and an appeals process for the mistakes. Every such system has a false positive rate and a false negative rate, and platforms choose which of the two errors they would rather make.
Why it matters: at billions of decisions a day, even a very small error rate means millions of wrong ones.
Read more: Automated Moderation Is Here to Stay—Accountability Must Keep Pace
Data broker
A company that buys, combines and resells information about people who are not its customers — location traces, purchase histories, inferred health conditions, public records. You have no relationship with it, no contract, and usually no idea it holds a file on you.
Why it matters: data handed over for one harmless reason ends up sold for a purpose you would never have agreed to.
Read more: Protect Your Privacy with California’s DROP Tool · The FCC’s Spam Call Proposal Is Just a Data Collection Scheme
Deplatforming
Removing an account, a channel or an entire service from a platform, so that the person or organisation loses the audience and the infrastructure it was using. Unlike a shadowban it is visible, and it is normally announced.
Why it matters: it is the bluntest instrument available, which is why both its use and its misuse are easy to see and argue about.
DSA — Digital Services Act
The EU regulation that sets out how online platforms must operate rather than what they must think: a formal notice-and-action route for reporting illegal content and appealing decisions, published transparency reporting, risk assessments, and access to platform data for researchers. The largest platforms carry the heaviest duties.
Why it matters: it is the first serious attempt to regulate the way platforms decide, not the outcome of each decision.
Read more: EFF and ARTICLE 19 Submission to the European Commission on the DSA Trusted Flagger Guidelines
End-to-end encryption (E2EE)
Only the devices at each end hold the keys, so the service carrying the message cannot read what it says. It protects the content; it does not hide who you spoke to, when, how often, or anything at all from someone holding your unlocked phone.
Why it matters: E2EE is necessary and not sufficient — most real-world exposure happens around the encryption, not through it.
Read more: Encrypted messaging without the myths
Facial recognition
Two very different jobs share one name. Verification (one-to-one) asks “is this the face that unlocks this phone?” Identification (one-to-many) asks “who is this face, out of millions?” — and that second question requires a database of everyone, and produces confident-looking mistakes.
Why it matters: the same two words in a law can authorise unlocking your own phone, or scanning a crowd for a face.
Read more: EFF, TEDIC and CEJIL Challenge Secrecy in the Use of Face Recognition
Geoblocking
Serving, restricting or blocking content according to the country your connection appears to come from, usually judged from your IP address. It is used for licensing deals, for legal compliance, and for censorship — and from the outside all three look identical.
Why it matters: “not available in your country” conceals who made the decision, and why.
Metadata
Everything about a communication except its content: who, to whom, when, for how long, from which device, from where. It is structured, cheap to store and easy to analyse at scale — three things that content is not.
Why it matters: content tells you what was said once; metadata tells you the shape of an entire life.
Shadowban
An umbrella term for reducing the reach of an account or a post without telling its author: nothing is removed, no notice arrives, simply fewer people ever see it. Platforms rarely use the word, preferring “reduced distribution” or “borderline content”, and it is hard to prove from outside because ordinary fluctuation looks the same.
Why it matters: an invisible penalty cannot be appealed, and often cannot even be confirmed.
Threat model
A plain answer to four questions: what are you protecting, who might want it, what can they realistically do, and what happens if they succeed. Security tools only make sense as an answer to a specific threat model; there is no such thing as “secure” in the abstract.
Why it matters: without one you buy tools for the wrong problem and feel safe anyway.
Read more: Encrypted messaging without the myths
Tor
A network that routes your traffic through several volunteer relays, each of which knows only the hop before and the hop after it, so no single point knows both who you are and what you are reading. It is slower than an ordinary connection, and considerably stronger against network observers than any single-provider VPN.
Why it matters: it is one of the few tools that also protects you from whoever is providing the tool.
Transparency report
A periodic publication in which a company states what it was asked to do and what it did: government requests, takedowns, account actions, security practices. A useful one gives numbers, categories, legal bases and a time series; a decorative one gives a single total with no denominator.
Why it matters: without comparable numbers, “we take your privacy seriously” is a slogan, not a claim you can check.
Read more: Most Smart Watches, Rings, and Bands Lack Basic Transparency Reports and Key Privacy Features
Missing a term, or think one of these is wrong? Tell us — we would rather fix a definition than defend it.
