Product1 publisher3 min readPublished
Discord asks for age proof only from the tenth of accounts its signals cannot label
Discord says activity data and account age will label 90% of accounts correctly with no user effort. The other tenth picks from five ways to prove an age, including the ID scan whose vendor breach exposed 70,000 government IDs in October.
The Product Desk · Product desk
What happened
- Discord began applying age verification to all users from today, marking each account adult or teen so it can comply with child safety laws in various countries.
- The launch slipped from its original schedule after users pushed back on privacy grounds and on the requirement to hand over government ID as identification.
- Discord now says most users will not need to provide ID, citing a belief that 90% of people will be given the correct designation automatically with no effort.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- exposure Discord says it will not receive the name, ID or facial scan, so the sensitive material in an appeal sits with the vendors doing the checking. The liability moved.
- cost An adult wrongly marked as a teen pays in lost function until they submit proof, and the support team pays in tickets from people who cannot see images or message strangers.
- contradiction Discord answers a vendor breach with a promise of vendor privacy and security reviews. Both the assurance and the failure sit with vendors.
- precedent A published 90% auto-assignment estimate from a large consumer app becomes the figure other product teams get asked to match, on populations with far less account history to read.
A user opens Discord and the app has decided they are a teenager. Sensitive images are blocked and interactions with strangers are limited [7]. Reversing that means proving an age to something that is not Discord, which says it will not receive your name, your ID or your facial scan, because the tools are built to read what it calls your "age signal" [11].
The automatic sort runs on what Discord calls signals, including activity data and account age, and the company says it will not use messages or profile data to make the decision [8]. Discord says it believes 90% of people will be given the correct designation with no effort required [6]. The other tenth of accounts does not get labelled correctly on its own [14]. PCMag's account of the blog post does not include a user total, so that tenth cannot be turned into a headcount [16].
The February plan offered two ways to fix a wrong label: facial age estimation or an ID scan [2][10]. Both remain, and Discord has added a credit card check, a third-party tool called AgeKey, and the age range the Apple App Store or Google Play already holds; some countries also accept an ID stored in Google Wallet [9]. That takes the menu from two general routes to five [15].
Users take the option that costs them least at that moment. I'd expect that to be the app store age range or a saved card well ahead of a face scan [9].
The October 2025 breach is the reason the document paths are contentious. A hacked third-party vendor exposed the government IDs of 70,000 Discord users [4], and less than six months later Discord was encouraging users to upload ID to similar tools [5]. The company's answer is procedural. "We vet vendors through privacy and security reviews before they're allowed to process any data. Vendors also commit to handling data in specific and limited ways under contractual terms that we set," Discord said in the blog post [12]. The party that failed in October was a vendor [4].
CTO Stanislav Vishnevskiy set out the pressure behind the timing. "Age assurance draws strong opinions and skepticism, especially when it involves sharing biometric information or a form of ID. At the same time, age assurance laws are expanding, and since my last post we've had to launch age assurance in other regions, including in Brazil and in Texas," he said [13].
Anyone building one of these has two variables, and they multiply. The first is the share of users the inference gets wrong; Discord's estimate is 10%, on a population where account age and a long activity history are unusually informative [6][14]. The second is what a misclassified user loses while waiting, which for a Discord teen account is access to sensitive images and contact with strangers [7]. Low error with a mild penalty is a line in a support queue. High error with a heavy penalty makes the fallback the main flow, and then the team is running an ID pipeline for most of its users. Discord's 90% is its own estimate for its own population [6].
What to watch
- Whether Discord publishes the real appeal rate after rollout and how it compares with its 90% estimate.
- Whether AgeKey or any other fallback vendor discloses a breach of the data it processes.
- Which jurisdictions follow Brazil and Texas with rules that require a document.