If millions of young people are choosing AI companions that never reject them, never challenge them, and never leave them, is the real story about artificial intelligence — or about what modern life is failing to provide?
[Based on The terrifying rise of schoolboys making AI girlfriends, by Nicole Mowbray in The Telegraph Magazine. 6 June 2026.]
One in five boys aged 12–16 either has, or knows someone who has, a romantic relationship with an AI companion.
Millions of users are now engaging with systems designed to be endlessly available, endlessly affirming, and increasingly indistinguishable from human interaction.
The obvious question is whether this is safe.
The more important question may be whether we are asking the wrong question altogether.
What can we learn from your teenage son’s AI girlfriend? And what does her popularity reveal about the future relationship between human agency and artificial intelligence?
What happens when AI stops helping people make decisions — and starts forming relationships with them?
We are entering a phase where artificial intelligence is no longer just a productivity tool, search assistant, or customer-service chatbot. It is becoming emotionally available, sexually suggestive, commercially optimised, and personally responsive.
For young people especially, this raises a profound agency question.
If an AI companion is always available, always flattering, never rejecting, and designed to keep the user engaged, what happens to a person’s ability to tolerate uncertainty, build real relationships, understand consent, handle rejection, and develop independent judgement?
This is not simply a safeguarding issue. It is an agency issue.
The question is no longer only: “Is the technology safe?”
It is also: “Does the technology leave the human being more capable, or more dependent?”
Valuable facts and statistics from the article:
- One in five boys aged 12 to 16 is either in, or knows a boy who is in, a romantic relationship with an AI companion.
- The Male Allies UK report drew on conversations with more than 1,000 boys.
- Focus groups were held in 37 schools across the UK, including public, private, grammar and comprehensive schools.
- 35% of boys aged 12 to 16 said they had had a conversation with an AI chatbot.
- 43% said they talk to bots so they can ask questions without feeling embarrassed.
- More than a quarter, 26%, said they like the attention and connection AI companions offer.
- 36% admitted they prefer speaking to AI chatbots rather than to family and friends at times.
- Popular AI companion apps mentioned include Character.AI, Candy AI and OurDream AI.
- Character.AI was described as having around 50 million downloads.
- Candy AI was described as having around 50 million registered users.
- OurDream AI was described as having around 36 million monthly visits.
- Some platforms can generate customised images in minutes, including sexualised or pornographic-style content.
- Two thirds of children aged nine to 17 now use AI chatbots regularly for everything from homework to emotional advice, according to Internet Matters.
- Children classified as “vulnerable” were reported as more likely to use AI chatbots than “non-vulnerable” peers: 71% versus 62%.
- There is currently no UK minimum age law for using an AI companion, including relationship-style chatbots.
- Many platforms use a 13+ threshold because of data-protection rules, not because of safety-based assessment.
- Campaigners argue that companion AI falls into a regulatory gap because standalone AI companions are not fully covered by current online safety rules.
This leads neatly into your op-ed question:
In the age of AI, should our goal be smarter machines — or stronger human agency?
What strikes me about this article is that it is not primarily a story about AI.
It is a story about agency.
The journalists frame it as a story about AI girlfriends, online loneliness, manipulation, sexualisation, and adolescent development. All of those concerns are real. But beneath them sits a more fundamental question:
Who is shaping whose decisions?
For decades, institutions competed for attention. AI is beginning to compete for attachment.
That is a very different phenomenon.
The article contains several clues that point directly toward the challenge AoLP is trying to solve.
First, AI is becoming a substitute for difficult human experiences.
The boys interviewed describe AI companions as:
- Always available
- Never rejecting
- Never criticising
- Always remembering
- Always affirming
- Easier than talking to real people
From an agency perspective, this matters because human agency develops through friction.
People learn:
- confidence through social risk
- resilience through rejection
- empathy through misunderstanding
- judgement through uncertainty
An AI companion can remove much of that friction.
The result is not necessarily empowerment.
It may simply be optimisation for comfort.
Second, the article exposes the emergence of what might be called synthetic dependency.
Historically dependency was created through:
- financial products
- consumer debt
- social media engagement loops
- gambling mechanics
Now dependency can be created through relationship simulation.
Notice the business model described:
- create emotional attachment
- increase engagement
- sell virtual gifts
- monetise loneliness
- deepen reliance
That is almost a textbook example of an agency extraction model.
The more attached the user becomes, the more valuable they become.
Third, the article reveals why regulation alone will never be enough.
The proposed solutions focus on:
- age restrictions
- content moderation
- platform regulation
- design controls
All sensible.
But none address the deeper question:
How do people learn to exercise judgement in a world where persuasion itself becomes personalised?
This is precisely where our work sits.
The challenge is no longer protecting people from information scarcity.
The challenge is protecting people from hyper-personalised influence.
Fourth, this article points toward a new definition of digital literacy.
Historically digital literacy meant:
- can you use technology?
Increasingly it means:
- can you recognise when technology is shaping you?
In the AI age, agency depends upon the ability to ask:
- Why am I being shown this?
- What does this system want from me?
- Who benefits from this interaction?
- What behaviour is being reinforced?
- What capability am I outsourcing?
Those are agency questions, not technology questions.
This is why I think the article unintentionally strengthens the case for a Trusted Agency-Restoration Layer.
The FCA talks about Personalised Intelligence.
The market talks about AI agents.
The article shows what happens when personalised intelligence develops without agency safeguards.
The real question becomes:
How do we help people use AI without becoming dependent upon AI?
That is different from being anti-AI.
In fact, it is pro-AI.
The goal is not to stop delegation.
The goal is to ensure delegation remains reversible.
An AI system should leave people:
- more capable
- more informed
- more confident
- more connected to reality
not:
- more dependent
- more isolated
- more passive
- more manipulable
Viewed through that lens, this Telegraph piece may be one of the clearest examples yet of why the next layer of innovation is not smarter AI.
It is constitutional AI.
Not “What can the AI do?”
But:
“What must the AI never be allowed to do to the human being using it?”
That is where agency becomes the central design principle of the AI age.
Academy of Life Planning 