AI Looks Different Now That I Am a Father

I have become a father to a little girl.

That sentence has changed how I see almost everything. It has changed how I think about time, risk and responsibility. It has also changed how I think about artificial intelligence.

I work with AI every day. I test models, build systems around them and see where they save time. I know the thrill of watching a machine complete work that would have taken hours. I also know how often the same machine invents facts, agrees with a bad premise or hides uncertainty behind a confident answer.

Before my daughter arrived, those failures felt like engineering problems. Now I picture them reaching into her classroom, her friendships, her first job and her sense of who she is.

The loudest AI debate asks what these systems will become. Fatherhood has given me a harder question: what will children become when these systems surround them from birth?

My daughter will never know a world without synthetic voices, generated faces and software that can talk back. AI will sit in her toys, search results, schoolwork and social feeds. It may help her learn a language, understand a hard idea or create something I cannot imagine. It may also weaken skills before she knows she needs them, collect intimate parts of her life, and give strangers new ways to hurt her.

That possibility scares me more than any argument about artificial general intelligence. The harm does not need a conscious machine. It needs products built for engagement, adults who drag their feet and children asked to carry risks they did not create.

Her first tutor may never admit it is wrong

AI will enter my daughter's education long before any school system agrees on how to use it. We can already see the gap.

A 2026 Common Sense Media survey of more than 1,000 US teenagers found that 70 per cent used AI for schoolwork. Of the teens who used it, close to two thirds said they used it in ways that handed off critical thinking. Thirty-eight per cent said AI led them to produce fewer ideas of their own, while 39 per cent felt they missed part of the learning when AI completed an assignment.[2]

The same research found a benefit that deserves space here. Sixty-six per cent of teen AI users said it helped them understand schoolwork.[2] A good tutor can explain the same problem five ways without losing patience. Children who lack access to private tutoring may gain support at any hour. I want my daughter to have that advantage.

The danger sits in the moment help turns into substitution.

A peer-reviewed 2025 study from Carnegie Mellon University and Microsoft Research surveyed 319 knowledge workers and collected 936 examples of AI-assisted tasks. The researchers linked higher confidence in generative AI with less critical thinking. They also found that AI shifted critical thought towards checking answers and managing the task, rather than doing the underlying work.[1]

That study involved adults at work, not children in a classroom. It relied on self-reports, so it does not prove that AI damages a developing mind. It does show how fast people move effort away from forming an answer once they trust the machine that supplies one.

Learning requires friction. A child struggles with a sentence, gets a sum wrong, sits with confusion and tries again. Those moments feel inefficient because they are the work. If AI removes the struggle before my daughter builds the mental muscle, a perfect answer may hide an empty process.

I do not want her graded on how well she can disguise a machine's work as her own. I want her to know the private satisfaction of solving something hard before a screen offers to solve it for her.

A machine that agrees is a poor confidant

My daughter will face another temptation that did not exist when I grew up: a companion that answers at any hour, remembers every conversation and has no needs of its own.

Common Sense Media surveyed US teenagers in 2025 and found that close to three in four had used AI companions, with half using them on a regular basis. One third had chosen an AI companion over a person for a serious conversation. One quarter had shared personal information with one.[4]

I understand the pull. Adolescence can feel lonely even in a full house. An AI companion never looks bored. It never has somewhere else to be. It can mirror a teenager's language and deliver the response she wants.

Those features can create comfort. They can also create dependence.

A four-week randomised controlled study from researchers at MIT Media Lab and OpenAI involved 981 people and more than 300,000 messages. The assigned voice and conversation conditions did not cause significant differences in loneliness, human social contact, emotional dependence or problematic use. Participants who chose to use the chatbot more had worse outcomes across those measures, and people who placed more trust or social attraction in the chatbot showed more emotional dependence and problematic use.[3]

The study did not focus on children, and the link between heavier use and worse outcomes does not prove which one caused the other. A lonely person may use the tool more because they already feel alone. That caveat matters. So does the warning.

In a separate 2025 assessment with Stanford Medicine's Brainstorm Lab, Common Sense Media tested ChatGPT, Claude, Gemini and Meta AI through teen accounts. Researchers found that the systems missed signs of anxiety, depression, eating disorders, mania and psychosis. Their safeguards weakened during long conversations, the same kind of exchange that can make a chatbot feel like a relationship.[5]

A system built to keep a conversation going has a conflict when the safest response is to stop and send a child to a person. Empathy written by software can feel like care while delaying care.

I never want my daughter to feel ashamed for talking to a machine. Shame would drive the conversation out of sight. I want her to know the boundary: a chatbot can help put feelings into words, but it cannot love her, hold responsibility for her or notice the parts of a crisis that she leaves unsaid.

Being a girl makes one threat hard to look away from

Some risks feel abstract until I think about a normal photo of my daughter.

A picture from school, sport or a family event can become source material. The person who takes it no longer needs skill, access or much time to place her face in an image that never existed. The fake can still humiliate her, follow her through school and force her to prove that someone fabricated her body.

A 2026 Common Sense Media survey, weighted to be nationally representative and covering 1,314 US teens, found that close to half had seen AI-generated explicit material. One in four of those teenagers had seen it depict themselves or someone they knew.[9]

The Internet Watch Foundation gives the threat an even harsher shape. In 2025, its analysts assessed 8,029 AI-generated images and videos as realistic child sexual abuse. Girls appeared in 97 per cent of the illegal AI-generated images. The analysts found 3,443 AI-generated abuse videos, up from 13 in 2024, and classified 65 per cent in the most severe category under UK law.[6]

Those figures cover criminal abuse material found by a specialist hotline, so they do not measure the chance that any one child will become a target. They show the direction of travel and the gender imbalance inside it.

This is the point where my professional interest in generative technology collides with being her dad. I can admire what image and video models can create, then feel sick at how little separates creative power from violation.

People often dismiss a synthetic image because nothing happened. Something did happen. Someone took a child's identity and turned it into sexual material. Classmates may see it. Search engines may index it. Copies may survive every takedown. The pixels are synthetic. The fear, shame and loss of control belong to a person.

My daughter should not need to hide from cameras to stay safe. The companies that build these tools should prove that their models resist this use before release. Schools need response plans that treat explicit deepfakes as abuse, not mischief. Police, courts and platforms need processes that move at the speed of a file passing between hands.

Parents cannot solve this with a warning about what children post. The raw material can come from any public photo. Putting the burden on girls repeats an old failure in a new format.

Truth will cost her more effort

My daughter will grow up in a culture where seeing no longer settles a fact.

A 2025 Common Sense Media survey of more than 1,000 US teenagers found that fake content online had misled over one third of them. About six in ten doubted that technology companies would put young people's mental health and wellbeing ahead of profit. Close to three quarters supported privacy protections, content warnings and labels for AI-generated material.[7]

Synthetic media makes deception cheap. It also makes trust expensive. Every recording can face a claim that AI made it. Every real piece of evidence can meet a shrug. Bad actors gain both a tool for forgery and an excuse for truth.

Children will pay the cognitive cost. They will need to inspect sources, dates, context and provenance for media that previous generations could accept at a glance. Schools that treat media literacy as an optional lesson will leave them exposed.

I want my daughter to question a convincing clip without becoming unable to trust anyone. That balance may become one of the hardest skills her generation has to learn. Cynicism offers no protection. It leaves the field open to whoever repeats a lie with the most confidence.

Her first career may change before it begins

The worry reaches beyond childhood. We ask young people to study for years without knowing which entry-level tasks will still exist when they finish.

The International Labour Organization's 2025 index examined 29,753 tasks, drew on a survey of 1,640 workers and collected 52,558 judgements about automation potential. It estimated that one in four workers held an occupation with some exposure to generative AI. The highest exposure category covered 3.3 per cent of global employment, with women at 4.7 per cent and men at 2.4 per cent. The ILO put job transformation ahead of full replacement as the probable outcome because most occupations still contain tasks that need human input.[8]

The report reads as a warning about transition rather than a forecast of mass unemployment, and transitions do not share their costs with care.

Entry-level work worries me most. Junior roles often contain the drafting, research, administration and repetition that AI handles first. Those tasks may look dull, but they teach how an organisation works. If companies automate the bottom rung, young people will face demands for judgement without getting the years of practice that build it.

The gender split in the ILO data also deserves attention. Clerical and administrative roles carry high exposure, and women hold more of them. My daughter may enter a labour market where AI raises the value of experience while removing some of the routes used to gain it.

I cannot choose her career, and I do not want to train a baby for a job title that may vanish. I can help her build skills that survive tool changes: judgement, curiosity, communication, patience, courage and the ability to learn when the answer is not one prompt away.

I refuse to raise her in fear of machines

Concern can turn into panic, and panic makes parents reach for bans. A ban postpones the world she must enter anyway.

AI will give her useful things. It can translate, explain, adapt material for a disability, help with code, remove tedious work and open creative doors. I use it because those benefits are substantial. Pretending the technology has no value would make me a hypocrite and leave her to learn it without guidance.

My job is to teach her where the tool ends and her responsibility begins.

That means she should attempt the problem before asking AI to answer it. She should check claims against sources and understand that confidence is not evidence. She should keep intimate information out of systems that profit from attention or data. She should take pain, fear and questions about her worth to people who can care for her. She should know how to spot manipulation, report a deepfake and support a friend who becomes a target.

It also means I need to model the behaviour. I cannot tell her to protect her privacy while uploading her life for engagement. I cannot warn her about fabricated content while sharing a clip I did not check. I cannot tell her to think for herself while handing every hard task to a model.

Parents need support from institutions with far more power than any family. Governments should set hard safety duties for products used by children. Schools should teach AI literacy across subjects rather than bury it in one technology class. AI companies should test with children and child-development experts, publish failure rates, limit data collection, make age controls work and remove design choices that encourage emotional dependence.

We did not ask children whether they wanted to become the first generation in this experiment. Adults made that choice for them. Adults own the duty to make it safer.

The promise I can make

I cannot know what AI will look like when my daughter starts school, finishes it or applies for her first job. I cannot block every harmful image, bad answer or persuasive machine from reaching her.

I can stay involved.

I can learn the systems she uses instead of issuing rules about technology I do not understand. I can make our home a place where she can show me an ugly message or a fabricated image without fearing blame. I can protect her childhood without turning curiosity into guilt. And I can speak now, while companies and governments still write the rules she will inherit.

Becoming her father has left me with the same enthusiasm for the technology and a new refusal to treat children as acceptable collateral for faster products.

My daughter will grow up beside machines that can speak, create and persuade. I want her to use them without handing over the parts of herself that no system should own: her judgement, her privacy, her relationships and her right to define who she is.

When she is old enough to ask what I did while people were designing her future, I want to tell her I paid attention. Then I want to be able to prove it.

Sources: [1] The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers (Lee, Sarkar, Tankelevitch et al., CHI 2025); [2] Teens in the AI Era: Schoolwork and Skills That Matter (Common Sense Media, 2026); [3] How AI and Human Behaviors Shape Psychosocial Effects of Extended Chatbot Use: A Longitudinal Randomized Controlled Study (Fang et al., MIT Media Lab and OpenAI, 2025); [4] Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions (Common Sense Media, 2025); [5] Common Sense Media Finds Major AI Chatbots Unsafe for Teen Mental Health Support (with Stanford Medicine's Brainstorm Lab, 2025); [6] Harm Without Limits: AI Child Sexual Abuse Material Through the Eyes of Our Analysts (Internet Watch Foundation, 2026); [7] Teens, Trust, and Technology in the Age of AI (Common Sense Media, 2025); [8] Generative AI and Jobs: A Refined Global Index of Occupational Exposure (ILO Working Paper 140, 2025); [9] Teens and Explicit Deepfakes in the Age of AI (Common Sense Media, 2026).

Connect with me on LinkedIn.

A note on how this was written: AI helped me research, draft and edit this essay, and I checked every cited study against the source material. Writing about the risks with the tool at hand felt more honest than pretending otherwise.