Tech Beetle briefing GB

Explainer: Age Assurance Tech Trial for Australian Under-16 Social Media Ban Reveals Challenges and Solutions

Essential brief

Key facts

Age assurance technologies can help enforce Australia's under-16 social media ban but errors are inevitable, especially near the age cutoff.

Multiple verification methods, including fallback options like ID checks, are needed to improve accuracy and reduce false negatives.

Biases in facial age estimation affect accuracy for non-Caucasian, female-presenting, and Indigenous users due to training data limitations.

Privacy concerns arise from some providers collecting excessive personal data in anticipation of regulatory needs.

No single technology suffices; a layered, privacy-focused approach combining various methods is recommended.

Highlights

Age assurance technologies can help enforce Australia's under-16 social media ban but errors are inevitable, especially near the age cutoff.

Multiple verification methods, including fallback options like ID checks, are needed to improve accuracy and reduce false negatives.

Biases in facial age estimation affect accuracy for non-Caucasian, female-presenting, and Indigenous users due to training data limitations.

Privacy concerns arise from some providers collecting excessive personal data in anticipation of regulatory needs.

Australia is preparing to enforce a ban on social media access for users under 16 starting December 2025.

To support this, a $6.5 million trial tested various age assurance technologies to verify users' ages and prevent underage access.

Conducted by the UK-based Age Check Certification Scheme (ACCS), the trial evaluated 60 technologies from 48 vendors, including facial age estimation, ID verification, parental controls, and behavioral age inference.

The trial found that while some technologies are effective and fast—most facial age estimation tests took under 40 seconds—errors are inevitable, especially for users within two to three years of the age threshold.

For example, a 17-year-old might be mistakenly identified as under 16, or a 14-year-old as over 16.

To address this, multiple fallback options such as document-based age verification are necessary to confirm age when initial methods are uncertain.

The report highlighted accuracy issues particularly affecting older adults, non-Caucasian users, female-presenting individuals near the age cutoff, and Indigenous Australians due to underrepresentation in training data.

Skin tone bias was noted as a significant factor, with darker skin tones reflecting less light and providing less detail for algorithms.

Privacy concerns also emerged, as some age verification providers appeared to prepare for potential regulatory investigations by collecting more personal data than necessary, raising the risk of privacy breaches.

Behavioral age inference, which estimates age based on user activity and connections, was seen as a low-friction tool to flag potential underage users and trigger further verification steps.

The trial also explored biometric methods like hand gesture and voice analysis, which showed promise but are not yet ready for widespread use.

Vendors reported tools to combat circumvention attempts such as AI-generated fake images and VPN usage, though these raise additional privacy and accuracy challenges.

Communications Minister Anika Wells emphasized that there is no "one size fits all" solution and that layered, privacy-conscious approaches are essential.

The trial's findings will inform Australia's approach to implementing its under-16 social media ban, balancing effectiveness with user privacy and fairness.