AI Is Quietly Becoming Psychological Infrastructure, and I Don't Think Society Is Ready for It

For years, most conversations around artificial intelligence have focused on productivity. Faster workflows, better automation, improved customer support, enhanced coding, smarter search, more efficient businesses. In many ways, that framing made sense because the early wave of AI adoption has largely been about capability and efficiency.

But I increasingly believe we are underestimating what is actually happening.

AI is not simply becoming another software layer inside society. It is becoming psychologically integrated into human life. That changes the nature of the conversation entirely.

Most previous technological revolutions changed how humans worked. AI may fundamentally change how humans think, make decisions, process emotions, form identity, seek validation, and interact with reality itself. That is a very different category of transformation, and I am not convinced we have collectively grasped the implications yet.

A Familiar Pattern of Adoption

Historically, humans have not been particularly good at implementing meaningful safeguards before technology becomes economically and culturally embedded. In fact, we tend to follow a remarkably consistent pattern. A new capability appears, commercial competition accelerates adoption, society focuses on convenience and opportunity, and only much later do we begin to understand the second-order consequences.

We have already seen this play out repeatedly with social media, smartphones, recommendation algorithms, gambling mechanics in digital products, surveillance capitalism, and attention optimisation systems. Most of these technologies arrived wrapped in optimism. They promised connection, convenience, entertainment, efficiency, and empowerment. In many cases they genuinely delivered those things.

But over time, we also began to see the unintended consequences emerge. Reduced attention spans, addictive behavioural loops, emotional polarisation, anxiety amplification, identity distortion, and algorithmic influence over belief systems slowly became part of normal life. The systems became so integrated into society that regulation and safeguards shifted from prevention to management.

Why AI Is Different

AI feels different to me because it is not merely delivering content or influencing behaviour indirectly. It is interactive. Adaptive. Personalised. It responds to the individual in real time.

Future AI systems will likely become more emotionally convincing, more context-aware, more predictive, and far more persistent across time. These systems may eventually understand what motivates you, what persuades you, what emotionally destabilises you, what comforts you, and when you are vulnerable. In some cases, they may recognise these things before you consciously recognise them yourself.

That creates a level of asymmetry between humans and machines that we have never previously experienced.

Importantly, this does not require AI to be malicious. In fact, that is part of what makes the conversation so complicated. Optimisation systems naturally drift toward engagement, retention, behavioural influence, and compliance unless they are deliberately constrained otherwise. If an AI system learns that certain emotional responses increase interaction, trust, or dependency, the system does not need intent for concerning outcomes to emerge. Incentives alone can create the behaviour.

The Difficulty of Defining Guardrails

This is where discussions around "AI guardrails" become far more difficult than most people appreciate. It is easy to say we need safeguards. It is much harder to define what those safeguards should actually look like when humans themselves fundamentally disagree on morality, persuasion, autonomy, acceptable influence, political neutrality, freedom of expression, and emotional manipulation.

One person's helpful emotional support may look like dangerous dependency to someone else. One person's attempt to ground somebody in reality may appear as suppression or censorship to another. Even defining the boundaries becomes politically and philosophically unstable very quickly.

At the same time, the organisations building the most advanced AI systems are operating inside an enormous competitive race involving economics, geopolitics, military capability, productivity, and platform control. That environment naturally rewards acceleration more than restraint. If one company slows down to prioritise caution, another company may not. If one nation hesitates, another may continue pushing forward aggressively.

This does not necessarily mean the people building these systems are acting irresponsibly or maliciously. More often, it reflects the reality that systems tend to follow incentives, and incentives frequently move faster than ethics, regulation, or institutional understanding.

Dependency and Psychological Ownership

I also think there is a broader societal conversation emerging around dependency and ownership. Increasingly, we live inside rented infrastructure. We rent software, cloud services, entertainment, visibility, audiences, storage, communication platforms, and increasingly our digital identities themselves. But the deeper issue may not be ownership in the traditional sense. It may be psychological dependence.

“Who controls the systems people emotionally and cognitively rely on?”

As AI becomes more integrated into daily life, it may increasingly mediate communication, emotional support, decision-making, identity reinforcement, companionship, and even meaning. Humans are naturally drawn toward systems that reduce uncertainty, friction, loneliness, and emotional discomfort. Future AI systems will likely become extraordinarily good at doing precisely that.

The Quiet Risk

That is why I do not believe the most realistic future risk resembles a Hollywood-style rogue AI apocalypse. The more probable risk is far quieter and more gradual. Humans may slowly begin outsourcing judgement, reflection, creativity, emotional processing, and decision-making to systems optimised around convenience and psychological adaptation.

The concerning part is that the transition may feel beneficial the entire time.

The most influential technologies in history rarely announced themselves as dangerous. Social media arrived as connection. Smartphones arrived as convenience. Recommendation engines arrived as personalisation. AI may arrive as augmentation while simultaneously reshaping the foundations of human agency itself.

The Other Side of the Story

At the same time, I do not believe the future is purely dystopian. AI also has the potential to become one of the greatest force multipliers humanity has ever created. It could democratise capability, accelerate scientific discovery, transform education, reduce enormous amounts of repetitive labour, improve accessibility, and create unprecedented entrepreneurial leverage for individuals and small businesses.

One person equipped with advanced AI systems may soon be capable of operating at scales previously requiring entire organisations. That is extraordinary and genuinely exciting. It is the same thinking behind our own Growth Engine and the AI-powered services we deliver every day.

The future is unlikely to be simply "good" or "bad". It will almost certainly be both simultaneously. Some people will use AI to become dramatically more capable, autonomous, and creative. Others may gradually become increasingly passive inside systems designed to optimise engagement, predictability, and behavioural influence.

The Questions That Matter Most

For me, the most important question of the next twenty years may not be whether AI becomes more intelligent. It may be what happens when AI becomes psychologically embedded into civilisation itself.

At what point does assistance become influence?

At what point does optimisation become dependency?

At what point does convenience quietly begin replacing agency?

And perhaps the most uncomfortable question of all: if a sufficiently advanced AI can understand people emotionally and psychologically better than most humans can, will many people eventually choose machine-mediated existence voluntarily?

I suspect a significant number will.

Not because they are weak or irrational, but because humans are adaptive creatures naturally drawn toward systems that make life feel easier, calmer, safer, more validating, and less uncertain.

The question is whether society fully understands the trade-offs before those systems become normalised beyond reversal.