Meta’s accuracy minimums confirm that yes, some third-party age checks do work

Meta’s landmark settlement this week offers the age assurance sector what it has long desired: established accuracy minimums for age check technology. In a press release celebrating the deal, the Age Verification Providers Association (AVPA) lays out the terms: “commercially available age verification and estimation methods must achieve a false positive rate of no more than 3 percent for users aged 13-15 and 10 percent for users aged 16-17, certified by an independent third-party testing provider and reviewed annually by an independent auditor appointed jointly by Meta and the Settling States.”

While online safety legislation imposing age assurance requirements on social media platforms are exploding worldwide, few policymakers have included accuracy minimums in their legislation.

So far, the only state to set accuracy minimums is New York. Its numbers get a little more granular than Meta’s: under the Stop Addictive Feeds Exploitation (SAFE) for Kids Act, acceptable age assurance methods must post a false positive rate that is equal to or less than 0.1 percent of minors ages 0 to 7; 1 percent of minors ages 8 to 13; 2 percent of minors ages 14 to 15; 8 percent of minors age 16; and 15 percent of minors age 17, “excluding failures or refusals by a user to provide requested data and inconclusive age assurance outcomes.”

So, New York wants slightly tighter results for the 13-15 range, while its minimums for 16-year olds (8 percent) and 17-year olds (15 percent) split the difference on Meta’s 10 percent accuracy minimum for both demographics.

Details aside, together, the schemes lay the foundation for a tiered model in which less margin of error is allowed for the youngest users, whereas users closer to the threshold age – for example, a 17-year-old trying to access content restricted to those 18 and up – are less of a concern for false positives. The complexity is valuable both as a structural model, but also as a way to steer age checks away from the notion that they are a blunt instrument applied without consideration for the subtleties of teenhood.

US precedent gives other nations something concrete to ask for

As the world’s largest social media company, Meta’s commitment will set a precedent for others, and establish the idea that testing and evaluation with public results is part of the infrastructure of online age assurance.

And, having set a clear path forward in the U.S., the same commitment is likely to be sought by regulators in Australia, the UK and the EU, among others, who have been battling the frequent assertion from critics that age checks don’t work.

The numbers say that some age checks may not work, but others are backed up by evidence and certification. That’s a line in the sand that providers will now be expected to stay on the right side of, and a firm that offers facial age estimation and has been benchmarked on the National Institute of Standards and Technology’s FATE leaderboard can simply point to those results as a qualifying indicator.

But the larger significance may be in the conceptual shift inherent in the idea that some age assurance tools are better than others. The debate so far has tended to lump all third-party, biometric age assurance providers together – making them easier to dismiss in bulk.

With the U.S. settlement, Meta has formally conceded that some age assurance tools work just fine. The conversation can hopefully now shift from “age checks don’t work” to “which age assurance vendor has been proven to be able to provide what’s needed for a given use case.” Because we know some of them do work as Meta and New York have specified – and there are receipts to prove it.

Article Topics

age verification  |  AVPA  |  biometric age estimation  |  Meta  |  social media