Discord Wants to Estimate Your Age Without Asking for ID
Discord is moving age checks into the background.
The platform began rolling out a new age-assurance system this week that uses machine learning to classify accounts as adult, teen, or unconfirmed based on account and behavioral signals. Discord says more than 90% of users should not need to complete a separate age check, while accounts the system cannot classify may be asked for additional verification.
The important shift is that age assurance can now happen passively, using information Discord already has about an account. CTO Stanislav Vishnevskiy said the system will not read private messages, voice calls, or profile content when making the estimate.
Announcing the feature, Discord’s CTO Stanislav Vishnevskiy addressed privacy concerns, noting that the tech would not read users’ messages, voice calls, or profile content to make the decision.
Vishnevskiy also announced that users can opt into other services they already trust to provide an age range, making Discord’s new age-check feature less intrusive than when the platform first announced it earlier this year.
How Discord’s age model works
Discord’s new age-assurance method aims to move away from conventional systems that explicitly ask users to submit valid proof of age, process and validate it, then assign a pass-or-fail result.
Touting the new system, Vishnevskiy said more than 90% of users may never see a prompt asking them to confirm their age. The company also says the underlying system is not trained on age documents
To do that, Discord says the system looks for usage patterns. It considers signals such as how long the account has existed, aggregated activity, communities joined, device and platform information, and the broader structure of its relationships with other users, servers, and games.
The strongest signal comes from what Vishnevskiy calls Discord Entity-Relationship Embeddings (DERE), a system that converts an account’s position in Discord’s social graph into a 32-dimensional numerical representation.
Because people of similar ages tend to have predictable similarities, the model can use those patterns to estimate whether an account likely belongs to the 13–17 or 18+ group without needing the content of those interactions.
Advertisement
Users can also rely on external age signals
Users who do not want to depend solely on Discord’s estimate may be able to provide an age range through services they already use.
Depending on the region, Discord says it can accept age information from providers such as Apple, Google, digital wallets or payment processors including Stripe.
According to The Register, even these come with privacy concerns, suggesting some users simply want the old way back.
The goal is to let Discord confirm an age range without collecting another copy of a user’s identity documents. That still raises questions about how much age-related information platforms should exchange, but it can reduce the need for Discord itself to hold sensitive verification material.
Accuracy remains an open question
One obvious question is accuracy. While Discord has positioned the system as highly accurate, systems like this are often susceptible to errors.
Discord has acknowledged that implementation problems will matter and has asked users to judge the rollout partly by how quickly it corrects mistakes.
There is also the question of aging out of a category. Someone classified as 17 today will eventually turn 18, and Discord has suggested that some form of recurring age assurance may be necessary. Details of how often those checks would happen remain unclear.
Could this be a new direction for the internet?
Discord’s rollout raises a question that now stretches well beyond Discord: is the internet slowly becoming age-gated?
The signs are increasingly visible, with governments worldwide pushing platforms toward age-assurance systems as part of efforts to keep children away from age-inappropriate content and services.
Advertisement
Discord’s own approach is notable because the platform has a particularly privacy-conscious and technically sophisticated user base. Yet, it has still moved toward automated age estimation rather than leaving age entirely to self-declaration. Vishnevskiy has also acknowledged that laws requiring age assurance are expanding, suggesting that platforms may have less room to avoid these systems as more jurisdictions introduce their own requirements.
The bigger concern is what happens if this becomes a normal gateway to the internet. Even privacy-preserving systems still need some way to establish whether a person belongs to one age group or another, potentially making age information a routine part of accessing platforms, content, and services.
What makes Discord’s approach notable is the attempt to reduce the amount of identity information users have to submit directly. If behavioral estimates and external age signals can reliably separate age groups without widespread collection of IDs or biometric data, other platforms may follow.
The larger question is what happens if age assurance becomes a routine gateway to online services. The debate will then be less about whether platforms should know a user’s exact age and more about how little information they can collect while still proving that someone belongs in the right age group.
Other news: Anthropic used roughly 950 Claude AI agents to analyze more than 200,000 enzymes, helping researchers uncover a previously uncharacterized biological system that could offer a glimpse of AI’s growing role in scientific discovery.
About Author
