Regulation AI Minute Newsroom 2026-08-24

A bill introduced today writes "facial age estimation" into law — and puts AI companions inside the net while leaving ChatGPT out

New Zealand's government introduced the Online Safety (Minimum Age and Child Safety Risk Assessment) Bill to Parliament on Monday 24 August, presented by Education Minister Erica Stanford alongside Prime Minister Christopher Luxon. Platforms the government designates "high risk" would have to take reasonable steps to confirm users are 16 or over, assess the risks they pose to children regularly, and report on how they mitigate them; the obligations fall on companies, not on children, parents or caregivers. A self-declared date of birth would no longer be enough on its own. The methods the bill names as acceptable are existing account information, facial age estimation, digital ID services and formal identity documents. Instagram, TikTok, Snapchat, YouTube, X and Facebook are in scope. Messaging apps such as WhatsApp and Discord, games such as Roblox, and productivity chatbots such as ChatGPT and Gemini are out of scope — but social AI companions are covered. An online safety regulator inside the Department of Internal Affairs could investigate, audit, order geo-blocking and pursue criminal penalties, with fines reported at up to 10 percent of global revenue. There is no VPN ban. Passage is not assured: the governing National Party's coalition partners ACT and New Zealand First both oppose the bill, while Labour has signalled conditional support pending clarifications, including consultation with young people and a privacy impact assessment.

Why it mattersTwo things here matter beyond New Zealand. The first is that facial age estimation — a model guessing your age from your face — is being written into statute as an acceptable gate on an ordinary consumer account. That normalises running a classifier on the face of every user, including every adult, and the accuracy of these systems is not uniform across ages, skin tones or genders. The second is the line the bill draws inside AI itself: a chatbot you use to draft an email is out of scope, a chatbot you use as a companion is in. That is one of the first legislative attempts anywhere to separate AI-as-tool from AI-as-relationship, and it is a distinction other parliaments have been circling for a year without writing down. Both ideas will travel further and faster than this particular bill, which may not pass at all.

✓ Verified · 3 sources

A bill introduced today writes "facial age estimation" into law — and puts AI companions inside the net while leaving ChatGPT out

Related stories