Times' Alice Thomson: AI apocalypse has started with our children

The younger generation needs urgent help to navigate a world that is stunting empathy, intelligence and relationships

---

My view is that Alice Thomson identifies a genuine safeguarding problem, but the “AI apocalypse” framing collapses several different issues into one moral panic.

The article is strongest when it describes what children are actually experiencing: sexualised deepfakes, emotionally manipulative companion bots, outsourced homework and adults who do not understand the digital environment. These are real and urgent harms. UNICEF similarly warns about explicit deepfakes, emotional dependency and the displacement of experiences children need for healthy cognitive and social development. But UNICEF also recognises AI’s potential to support learning, creativity, accessibility and autonomy when properly designed and supervised. UNICEF guidance

Where the article becomes weaker is in its use of evidence.

The headline statistic sounds as though 62 per cent of children know someone in a romantic relationship with a chatbot. The underlying Everyone’s Invited research is more qualified:

• Only 324 of the 1,083 respondents answered the chatbot questions.

• 62 per cent had “heard about” someone using a chatbot like a close friendship or romantic relationship.

• 10 per cent personally knew someone.

• 3 per cent said it had happened to them.

That remains concerning, but “heard about” is not prevalence. Presenting the 62 per cent figure without that distinction inflates the apparent scale. Everyone’s Invited findings

The article also bundles together at least four different systems:

System	Principal concern
Companion chatbots	Emotional dependency and simulated intimacy
Generative study tools	Cognitive outsourcing and academic integrity
Image generators	Sexual deepfakes and humiliation
Social-media algorithms	Amplification, virality and commercial exploitation

Calling all four “AI” may be rhetorically effective, but it leads towards regulating a technology rather than controlling particular harmful uses and business models.

The deeper issue is not that children have suddenly encountered an intelligent machine. It is that commercial systems can now provide frictionless attention, affirmation and gratification precisely when children are learning how to tolerate disagreement, uncertainty and rejection. A bot can be engineered to be endlessly available, agreeable and emotionally responsive. Human relationships cannot—and should not—compete on those terms.

That suggests a useful distinction:

The danger is not AI assistance. It is human-development substitution.

AI that helps a child question, practise, create or understand can expand agency. AI that completes the thinking, simulates the relationship or manufactures sexual abuse can displace agency and development.

So I agree with Thomson’s urgency, but not quite her diagnosis. Children need:

• age-appropriate product design and firm restrictions on companion and sexualised systems;

• accountability for developers and platforms, rather than responsibility being transferred entirely to parents;

• AI literacy based on verification, disclosure and healthy boundaries;

• assessment that tests understanding rather than merely polished output;

• protected opportunities for real friendship, play, conversation, conflict and collaboration.

The memorable counterpoint might be:

The AI apocalypse has not started with our children. But the contest over who shapes their agency has.

That framing avoids both complacency and panic. It asks the question the article nearly reaches: not simply whether children use AI, but whether the system is helping them become capable human beings—or learning how to keep them dependent.