It’s time to mandate age checks on AI chatbot users to protect those under 18

SINGAPORE - A new variant of ChatGPT for teenagers with targeted safeguards sounds like a win for youth online safety.

ChatGPT for Teens, rolled out from Aug 19, aims to filter out harmful content that promotes self-harm, violence, eating disorders, dangerous activities and explicit sex.

To achieve this, young users are prompted to declare they are under 18 during account registration, which is voluntary.

Registered users that have been detected by OpenAI’s system to be under 18 will also automatically be classified as a teen account.

But for such safeguards to reach all underage users, AI platforms must move beyond systems where creating an account is optional and where the age entered is never checked.

Since generative AI chatbots entered mainstream consciousness in 2022, they have quickly evolved into a useful tool for both leisure and work purposes.

But they have also brought about a new wave of online harms that range from exhibiting sycophantic behaviour that reinforces negative beliefs, to sexually explicit chats with minors, to providing suicide advice.

Earlier in January, xAI’s Grok chatbot came under fire for acceding to user requests to generate non-consensual, sexual content en masse on social media platform X, where Grok is embedded. Following the backlash, X restricted users from editing images of real people in revealing clothes, and further limited image generation and editing features to paid subscribers.

Several suicides have also been linked to AI chatbots, including those involving children.

In 2024, 14-year-old American Sewell Setzer III killed himself after becoming obsessed with talking to an online character created by Character.AI.

In a lawsuit filed in Florida against the AI chatbot start-up, his mother accused the firm of creating a product that targeted her son with “anthropomorphic, hypersexualised and frighteningly realistic experiences” that ultimately resulted in his desire to no longer live outside the fictional world that the chatbot had created.

Such reported cases highlight the severity of online harms that AI chatbots can pose, and that measures to curb harm should not be left as a good-to-have for artificial intelligence companies and users to implement and opt in as they see fit.

A framework already in use in Singapore provides a glimpse of how AI chatbot operators and toxic content produced can be effectively reined in.

Since April, app stores are required by the Infocomm Media Development Authority to prevent Singapore users under the age of 18 from downloading inappropriate apps, such as those providing dating services or with sexual content.

Users’ age can be determined by checking government-issued identity documents or credit card information, scanning faces or analysing online usage patterns.

There are plans to extend age checks to social media services. Particularly, unmanaged direct messaging from strangers and video auto-play are features of concern.

Increasingly, AI chatbots have also been embedded in social media and direct messaging services, blurring the lines between these service categories.

Lim Sun Sun, Lee Kong Chian Professor of Communication and Technology at Singapore Management University, argued for regulatory intervention such as mandatory age checks to be extended to all emerging tech, including AI chatbots that young people access.

“This would go towards securing greater ownership by AI companies in their responsibility towards younger users,” said Lim, who is also SMU’s vice-president of partnerships and engagement.

Without a consistent way of implementing age checks, users will simply migrate to chatbots with the least restrictions and the whole effort to protect minors risks being defeated.

Currently, AI firms do not have a consistent approach in the way they protect the young.

So far, Anthropic is the strictest, requiring every user to register an account. It routinely analyses user activities to determine if they are under 18, before disabling these underage accounts.

Similar to OpenAI, Google has safeguards for registered users under 18 by disallowing Gemini from claiming to be a person, having human emotions, or role-playing as a harmful character.

But a limitation is that users do not need a Google account to start a conversation with Gemini.

Character.AI – a platform that allows users to converse with fictional characters – has banned open-ended chats for under-18 users since 2025. This was done to tackle concerns from parents and safety experts about the unhealthy interactions teens were having with its chatbots.

Age checks protect minors on AI chatbots, but harmful interactions impact everyone.

This is why Zhang Renwen, Nanyang Assistant Professor at Nanyang Technological University’s Wee Kim Wee School of Communication and Information, proposed that youth safety filters should be the standard default for all users.

“Having built-in safety features that reduce harms for all users is still non-negotiable.”

Helplines

Mental well-being

• National Mindline: 1771 (24 hours) / 6669-1771 (via WhatsApp)

• Samaritans of Singapore: 1-767 (24 hours) / 9151-1767 (24 hours CareText via WhatsApp)

• Singapore Association for Mental Health: 1800-283-7019

• Silver Ribbon Singapore: 6386-1928

• Chat, Centre of Excellence for Youth Mental Health: 6493-6500/1

• Women’s Helpline (Aware): 1800-777-5555 (weekdays, 10am to 6pm)

• The Seniors Helpline: 1800-555-5555 (weekdays, 9am to 5pm)

• Tinkle Friend (for primary school-age children): 1800-2744-788

• Mindfull Community HQ (for caregivers, families and persons in recovery): 6460-4400 (weekdays, 9am to 5pm)

Counselling

• Touchline (Counselling): 1800-377-2252

• Touch Care Line (for caregivers): 6804-6555

• Counselling and Care Centre: 6536-6366

• We Care Community Services: 3165-8017

• Shan You Counselling Centre: 6741-9293

• Clarity Singapore: 6757-7990

• Care Corner Counselling Centre: 6353-1180