Bias in age-guessing software threatens rollout of Australia’s teen social media ban

A government-commissioned report in Australia has said that selfie-based age-guessing software could help enforce the upcoming ban on social media use by people under 16, Reuters reported. However, the study also found ‘unacceptable’ levels of inaccuracy for some groups, raising concerns ahead of the December rollout.

The report, released on Monday, reportedly said photo-based age estimation tools were generally accurate, fast, and designed to protect privacy. But accuracy dropped for users close to the age minimum of 16. The findings suggest uneven outcomes: Caucasians slightly above the age cut-off may face few issues, while non-Caucasians and teenage girls could experience more errors.

Social media platforms such as Meta’s Instagram and Alphabet’s YouTube are required under the new law to show they are taking reasonable steps to block users under 16. Failure to comply could lead to fines of up to A$49.5 million ($32 million), the media reports noted.

The trial showed strong accuracy for people over 19, but users within three years of the 16-year threshold were in what the report called a ‘grey zone’ of higher uncertainty. Sixteen-year-olds had an 8.5% chance of being misclassified as underage and could be asked to provide extra proof, such as ID-based verification or parental consent.

The study also found lower accuracy for older adults, non-Caucasian users, and female-presenting individuals near policy limits. Experts told Reuters that the results raise questions about whether the ban can be implemented effectively within three months.

The media report quoted Communications Minister Anika Wells saying the report showed there is no single solution to age verification, but added that “this trial shows there are many effective options and importantly that user privacy can be safeguarded.”