With governments increasing pressure on platforms from social media to adult sites, firms offering AI-based selfie age checks are experiencing strong growth.
The technology offers quick, accurate enforcement of measures such as Australia’s under-16 social media ban, which takes effect on December 10.
For users, the age-check process is simple: you snap a straight-on selfie with your phone or computer, and the system delivers a result in under a minute.
“We estimate you’re 18 or older,” a pop-up on the Roblox site announced.
Mannequins and Masks for AI Testing
Inside Yoti’s spacious London headquarters, mannequin heads—some wearing wigs or masks—sit along a windowsill for testing purposes.
You want to rewrite the sentence using active voice rather than passive or indirect phrasing. Here’s a clear active-voice version:
Over time, “the algorithm became highly skilled at spotting patterns and determining that ‘a face with these features looks about 17 or 28’… it just got very good at estimating age,” explained CEO Robin Tombs.
The startup now conducts about one million age checks per day for major clients including Meta, TikTok, Sony and Pinterest, according to Tombs.
Yoti became profitable this year after generating £20 million ($26 million) in revenue in the 12 months to March, and it expects sales to rise by 50% in the current financial year.
Dozens of Competitors Join the AVPA
Rivals are thriving as well. Thirty-four firms—such as Persona, Kids Web Services, K-id and VerifyMy—are listed as members of the Age Verification Providers Association (AVPA).
Back in 2021, the AVPA projected that the sector could bring in nearly $10 billion annually across OECD countries during 2031–36, though it has not updated those forecasts.
AVPA director Iain Corby remains cautious about predicting the future. “There are so many variables. Regulation is evolving quickly, and so is the technology,” he told AFP.
AI-based age checks have also sparked concerns about bias and privacy risks. Such tools can be “intrusive or potentially harmful to people’s private lives,” said Olivier Blazy, a cybersecurity professor at the French engineering school Polytechnique.
Much depends, he noted, on what data websites share with their age-verification partners.
“The ecosystem is leaning heavily toward AI right now, but we may see a shift in the coming years to better safeguard user privacy,” Blazy said.
Blazy noted that these systems still have vulnerabilities, pointing out that “ordinary makeup can make it fairly easy to appear older or younger.”
Bias Concerns Persist in AI Age-Verification Systems
Others have raised concerns about biases in age-verification algorithms, which tend to be less accurate when assessing non-white faces. An independent Australian review found that “underrepresentation of Indigenous populations remains an issue,” though it added that “vendors are beginning to address” the problem.
Yoti CEO Robin Tombs admits that the company had less training data for certain age ranges and skin tones. Still, he maintains that their technology can detect attempts to manipulate appearance through “false accessories, makeup, and similar tricks.”
On privacy, Tombs said Yoti’s system deletes all data immediately after completing its analysis.
Platforms using selfie-based age checks can also fine-tune the tool’s thresholds, allowing for a buffer. For instance, sites with an age limit of 18 often require the algorithm to confirm that a user looks at least 21.
Anyone who falls into the uncertainty zone may be prompted to use a more traditional method—typically taking a photo or scan of an official ID.
Read the original article on: Techxplore
Read more: Chinese Robot Breaks 106 km World Record in Tech race