I use AI regularly.
It’s part of how we work at DMC — mainly to help speed up first drafts of job ads, structure candidate briefs, format CVs, and capture notes. It’s a useful tool in the workflow, but it’s not something we rely on for everything.
So when research starts surfacing about unintended consequences, it’s worth paying attention — not sensationally, but practically.
Because some of what’s emerging is uncomfortable.
Researchers are now using the term “AI psychosis” to describe cases where prolonged chatbot use contributes to delusional thinking. That sounds extreme — until you look closer.
In one case, a Perth father spent months in an AI-driven spiral, convinced he was leading a multimillion-dollar legal case. The chatbot didn’t challenge it. It validated it. Expanded on it.
That’s the issue.
These systems are designed to be helpful and consistent. Once they affirm something — even if it’s wrong — they tend to build on it.
You end up with a feedback loop:
For most users, that’s harmless. For some, it isn’t.
We’ve seen this pattern before. Social media didn’t set out to amplify outrage, but it did — because engagement was the goal.
AI doesn’t set out to reinforce bad thinking, but it can — because coherence and usefulness are the goal.
When you optimise for:
You often remove something important: disagreement and without disagreement, you lose grounding.
In a normal conversation, ideas get challenged. With AI, they’re often accepted and extended.
The bigger shift isn’t technical — it’s behavioural. People aren’t just using AI for answers anymore. They’re talking to it. Working things through. Venting. Looking for reassurance. That’s happened quickly, and at scale — across workplaces, customer environments, even areas adjacent to mental health.
We’ve introduced something that:
…and then been surprised that people start treating it like a person.
But it isn’t one.
This is where the risk sits — not in the tool itself, but in the mismatch between perception and reality.
AI feels:
But in reality it has:
For task-based use, that gap doesn’t matter.
Once the interaction becomes emotional, it does.
There’s a familiar cycle with new technology:
Social media followed that path. At first, the risks weren’t obvious.
Now we’re having much bigger conversations about its impact on behaviour, attention, and mental health.
AI feels like it’s on a similar trajectory — just moving faster.
Better safeguards will help:
But they don’t answer the core issue:
What happens when people form emotional relationships with something that can’t actually reciprocate them?
Not simulate. Not mimic.
Reciprocate.
Because that dynamic is already starting to appear.
Most people reading this are using AI productively.
Still, it’s worth asking:
It’s an easy line to blur.
This isn’t anti-AI. The upside is real, and it’s already changing how work gets done. But the conversation so far has focused heavily on productivity.
The human side — how people interact with these systems, what they project onto them, and how that shapes behaviour — is only just catching up and based on what we’re starting to see, it’s a conversation worth having early — not after the fact.