This talk is part of the TrustCon 2024 Child Safety Lightning Talk session. Lightning talk sessions are 50 minutes long, featuring a series of short 5-7 minute talks back-to-back in the same room, followed by Q&A at the end.

Age assurance is already being deployed at scale to keep children safe online. In the UK, the Online Safety Act mandates its use to prevent children from encountering harmful content. While each kind of age assurance comes with trade-offs, biometric age estimation technologies are accessible to individuals without identity documents, minimise friction, and can be performed on-device, reducing the impact on privacy. But with generative AI making convincing facial filters easily accessible to children, what role should biometric age estimation play in Trust & Safety? This lightning talk presents the existential risk AI-facilitated presentation and injection attacks pose to age estimation technologies and introduces ongoing research aiming to mitigate this risk. It highlights that for age estimation technologies to be a proportionate intervention, it is not enough to limit their impact on user privacy. They must also be effective in an online world where generative AI is proliferating.