Age assurance in the UK and Australia: what to build once
Build one way to run an age check. Decide separately what each service does with the result. A publishing site may need to make a decision before showing content, while a social platform may need to decide whether someone can keep an account. That difference changes where teams check age, what they test, and what records they keep.
In this article
Start with the regulated decision
This comparison covers two specific cases: UK Part 5 services that publish their own regulated pornographic content and covered Australian social media platforms. The research was checked on November 20, 2026.
In the UK, Part 5 of the Online Safety Act 2023 requires covered services to use highly effective age verification, age estimation, or both so children cannot normally encounter regulated provider pornographic content. The duty came into force in January 2025. Ofcom’s guidance describes four criteria for effectiveness in practice: technical accuracy, robustness, reliability, and fairness. Section 81 also requires a written record of the methods and how they are used, and how the provider considered protecting UK users from breaches of applicable privacy law. It does not require keeping each person’s identity documents. [2] [1] [6]
In Australia, section 63D of the Online Safety Act 2021 requires covered platforms to take reasonable steps to prevent people under 16 who ordinarily live in Australia from having accounts. The duty took effect on December 10, 2025, and applies to existing accounts as well as new sign-ups. A platform collecting government-issued identification or using an accredited Digital ID service for that purpose must offer a reasonable alternative that uses neither. Section 63F(3) also requires entities to destroy personal information collected for this purpose after using or disclosing it for the purposes for which it was collected. [3] [4] [7]
What teams can build once
A shared integration can call approved check providers and report three outcomes to the product: a usable result, an inconclusive result, or a check that could not finish. Keep the method and the threshold it was used for part of the decision. A result that shows someone is at least 16 does not tell the service that the same person is at least 18. Each service decides what assurance it needs and what to do when the check falls short.
Teams can also share tools for measuring completion and errors. Start with aggregate counts and ask what any proposed personal field would help the team decide. Keep reporting about how the system performed separate from data collected to make a decision about a particular person. The joint Ofcom and ICO statement explains how online safety and data protection duties interact in the UK and emphasizes proportionate handling of personal information. [6] [2]
What to configure for each service
Imagine an operator with two products in scope. For its UK publishing site, the team maps ways a visitor might reach regulated content, including direct links, and makes sure the check happens before the content is visible. For its covered Australian social platform, the team looks at both sign-up and existing accounts. It also designs a way for someone who was restricted by mistake to request a review. The teams can reuse the check provider integration while setting different rules about access and account status. [2] [1] [3] [5]
Keep a short decision record for each service and market. Write down what triggers the check, the applicable age threshold, what counts as enough evidence, and what a person can do when the check is inconclusive or wrong. If an Australian platform uses government-issued identification or an accredited Digital ID service for the minimum-age duty, test that the alternative is practical to find and complete. Put the law, guidance, and last checked date next to the decision so the next team can see what could change. [3] [5]
Draw the data lifecycle as part of that record. The UK team can document how it chose and uses the process without collecting identity documents just for that record. The Australian team should trace where personal information used for the account duty is copied, including by providers, and how it will meet the destruction duty after the information has been used or disclosed for its collection purposes. Treat a retention setting as a decision the team must justify for its own use and obligations. [3] [7] [2]
Test errors at the threshold
In May 2024, NIST evaluated prototype facial age estimation and verification software using archived photos. Performance varied by factors including algorithm, age, and image quality, and the order of the prototypes could change with the chosen metric. The study gives teams a reason to test their own journeys near the actual threshold. It does not predict how well a specific provider will perform on a live website. [8]
For example, in a hypothetical test of an age-16 decision, 8 of 100 participants under 16 might be allowed through and 12 of 100 eligible participants might fail the first attempt. Those are two different rates, 8% and 12%. Report them separately and show how many eligible people complete an alternative path. A single overall accuracy number hides who is blocked or admitted incorrectly.
Test the product after the check as well. Can the UK publisher keep content protected when a check is unavailable? On the Australian platform, can an eligible user find the review route and can the platform apply a corrected decision to the account? Test a range of devices and capture conditions, and show how many people were in each group so readers can judge when results are too sparse to generalize. [1] [5]
A rollout checklist
Before rollout, ask the product, engineering, and privacy teams to review the same record. These are design checks teams can adapt for each covered service.
- Record the covered service, jurisdiction, source, effective date, decision point, and policy owner. [2] [4]
- Test the alternative path when government-issued identification or accredited Digital ID is used for the Australian minimum-age duty. [3]
- Map which personal information is collected for the Australian duty, where copies go, and how the destruction requirement will be met. [3] [7]
- Keep a UK Part 5 record of the methods used, how the service uses them, and how it considered applicable privacy law. [2]
- Test both sides of the age threshold. Report underage acceptance, eligible-user rejection, and completion through any fallback separately. [8]
- Test how a user reaches a review after an incorrect restriction and how a correction reaches the account system. [5]
Sources
- Age assurance duties under the Online Safety Act Ofcom
- Online Safety Act 2023, section 81 UK legislation register
- Australia’s Online Safety Act 2021, Part 4A Federal Register of Legislation
- Social media age restrictions eSafety Commissioner
- Social media minimum age FAQ eSafety Commissioner
- Joint statement on age assurance Ofcom and the Information Commissioner's Office
- Privacy Guidance on Part 4A of the Online Safety Act 2021 Office of the Australian Information Commissioner
- Face Analysis Technology Evaluation: Age Estimation and Verification, NISTIR 8525 NIST