An Australian government-commissioned report has found that selfie-based age estimation software could be used to enforce a proposed ban on teenagers under 16 using social media, but it also revealed significant accuracy issues that raise concerns about its planned December rollout. The report concluded that while such technology is generally fast, accurate, and privacy-respecting, it exhibits “unacceptable” levels of inaccuracy for certain demographic groups, particularly those near the minimum age threshold.
The findings indicate that the user experience will be highly uneven once the ban takes effect. The software showed high accuracy for Caucasians over the age of 19, but performance worsened significantly for non-Caucasian users and teenage girls. For 16-year-olds specifically, there is an 8.5% chance of being incorrectly identified as underage, which would force them to undergo supplementary verification methods like providing ID or obtaining parental consent.
The report places immense pressure on social media platforms like Meta’s Instagram and Alphabet’s YouTube. Under the new law, these companies must demonstrate they are taking “reasonable steps” to block underage users by December or face fines of up to A$49.5 million. The trial identified a “grey zone” of higher uncertainty for users within three years of the age cut-off, complicating the compliance process for tech giants.
Social media experts have expressed deep concern over the tight three-month timeline, given the variations in accuracy across different user groups. Justine Humphry, a media researcher specializing in online safety, stated that such variation is “concerning” for a system that needs to be robust by year’s end. In contrast, Communications Minister Anika Wells pointed to the report as proof that effective, privacy-protecting options exist, despite there being no one-size-fits-all solution.
