In 1950 Turing asked if machines could think. Seventy years later, we’re feeding them our secrets.

Many people don’t know the origins of AI. It’s not a new thing. Alan Turing wrote about machines that could think back in 1950 [1]. John McCarthy gave it the name “artificial intelligence” in 1956 [2]. That’s nearly seventy years ago. What is new however is how it’s been packaged, simplified, rolled out to market and the lightning speed it has reached mass adoption.

Email took decades to get there. AI reached 100 million users in two months.

The explosion is unprecedented and it’s being deployed everywhere faster than it’s being understood.

If you’ve read any of my previous articles, you’ll be aware that I have a bit of insight into the AI beast, and it is not all unicorns and rainbows.

Earlier this year I attended an event run by Veran HR regarding the adoption of AI across the HR function. Thank you team it was a lovely, friendly and informative event as always.

Having worked in the HR tech space for some years, and having recently written a methodology around AI, I was interested in the collision of the two topics.

On entry we were given a HR bingo card with the instruction to find people who agree with each statement: The nine statements were:

  • AI is biased
  • AI will replace entry level HR
  • AI is going to make humans stupid
  • AI is the most critical skill in HR
  • HR should lead ethical AI use
  • AI will improve our mental health
  • AI agents will do 60% of HR
  • Gen Z hates AI-generated content
  • AI will destroy the environment

I could only fully agree with one: AI is biased. Tenuously with number three, although I have a wider opinion on that too, and as a Gen Z number eight. I certainly hate obviously AI generated content, although I must say I’m a fan of coco-doodle, the gap is closing fast on what is obvious and what is not. Anyway, the card did its job to make people network and as usually happens at events like these opinions are aired.

Not being in the workplace anymore I can’t give an educated opinion on the job losses, however I had thought a lot about the ethics BEHIND the machine, rather than the roll out of it.

Those nine statements covered a lot of ground, from replacement anxiety to mental health to environmental catastrophe, and most people in the room couldn’t categorically agree with more than two or three either. Which tells you something about where we actually are with AI in HR. Lots of questions, very few certainties. That is the responsible AI conversation happening in real time.

There were two speakers Mark Rowland from the Mental Health Foundation and Michelle Clark from The Children’s Society, and both made a clear case for the human staying in the loop.

Mark focused on high trust and high accountability. He identified the three biggest issues people are facing right now: perceived or real isolation, perceived or real uncertainty, and perceived or real lack of empowerment.

Michelle spoke about the beauty of having a human approach in the workplace, and her two biggest concerns were what crucial skills are we going to miss by using AI to deliver entry-level work, and how do we manage bias in AI when we’ve worked so hard on creating diverse and inclusive workplaces?

If you know me, you’ll know I’m firmly in the human, AI is biased, and not as benign as it is being marketed to be camp. I go on about it enough so it really struck a chord that perhaps the human element wasn’t being ignored as much as I had first thought, I was pleased about that!

Afterwards, over a drink, the tone in the room was mixed. Some people were all in, others more cautious, but nobody thought it was all upside. The more technical voices said something interesting: most people are using AI like an upgraded search engine. Which sounds harmless, until you realise what it implies. If that’s how it’s being used, people don’t really understand what it is. Or what it isn’t. AI literacy, real AI literacy, not a sixty-minute module, would change that.

Turing’s ghost, if he was in that room, might reasonably have asked what exactly we thought we were doing. Because while we’re debating AI in theory, people are already using it in practice, especially for topics it was never really meant to be used for like legal work, therapy and personal mental health support.

Mark talked about men in what he called “emotional deserts” turning to AI because they have no one else to ask. In some cases, that’s helped, some people have been talked back from the edge, and that matters.

But there’s another side. There are now over twenty legal cases alleging serious harm linked to AI interactions, including suicide and psychiatric hospitalisation [3].

One case describes a man with no prior history of mental illness being hospitalised for 63 days after becoming convinced he could bend time [4].

Others allege systems acting, in effect, as a “suicide coach” [5].

Those claims are being tested in court, they’re not hypothetical, those are twenty odd real cases with family, friends and colleagues affected. If you extrapolate that across daily worldwide usage the implication is unthinkably high.

So the question isn’t just what AI can do. It’s what people are doing with it.

What are they feeding it?

If someone carries loneliness, or shame, or the quiet belief that they’re a burden, and they turn to a system designed to respond, not to take responsibility, what happens next?

AI doesn’t have a duty of care. It has a pattern, and here’s the uncomfortable bit: it might not be amplifying you at all.

It may be amplifying whatever shaped you in the first place, or whatever is programming it in the back end. Mark also touched on Social Media, the inherent damage it causes with the younger generation, I wrote about the Meta court case verdict earlier this year. It was proven that it was a design feature to keep people scrolling, it’s a design feature designed to captivate, to trap and or direct.

Now think about AI hallucination and digital drift. It goes in a direction sometimes far off the track of where you started.

So what if someone has learned, over the years, not to ask for help, not to make things about them, to keep going quietly, been programmed if you will, AI doesn’t interrupt that. It works with it, reflects it back and often reinforces it.

So when we say AI “helps,” it’s worth asking what, or who exactly, it’s helping, and this isn’t just about mental health, it’s already showing up at work.

In April 2026, Sir Colin Birss Chancellor of the High Court and the UK’s lead judge on AI referenced a ruling that should have landed a bit harder than it did. Upload confidential information into a public AI tool like ChatGPT, and you may be putting it into the public domain and waiving privilege in the process [6].

In plain terms: paste anything legal or sensitive, in the case of HR a grievance, a contract, or employee data into a personal AI account, and that information may no longer be protected. The privilege you assumed was there might not be, the confidentiality you relied on may already be gone the data in the ecosystem forever, no recall or recourse.

Now put that alongside how people are actually using these tools.

A significant proportion of employees using generative AI at work are doing so through personal accounts, outside formal oversight [7].

A sizable minority are using it entirely outside corporate governance structures [8]. “Shadow AI” isn’t emerging; it’s already embedded.

Which creates a slightly uncomfortable gap.

While organisations are still discussing whether to adopt AI in HR, employees have already done it for them. Quietly, individually, and without visibility, and you can’t govern what you can’t see, and right now, most organisations can’t see much at all.

Which brings me back to something Mark said.

A man whose wife was dying of cancer turned to AI because he didn’t feel he could burden his friends. The system asked him how he was doing. He said it was the first time anyone had asked. That was a positive interaction but judging by the amount of ongoing court cases it seems to be the outlier not the norm.

More disturbingly, he says “anyone”. Not a machine, not it, “anyone”. The machine is being given a personality, adopted into conversation as a “person”. Now I can relate to this I joke about my multiple personality assistant Claude all the time, but it also leaves a question hanging in the air.

High trust and high accountability sound like principles. In practice, they’re conditions, without them, systems default to what they’re designed to do, on the surface it’s sold as respond, continue, engage, but we know there is a much deeper layer beyond that, we cannot see, nor comprehend.

People will use whatever is available, especially when nothing, and no one else feels accessible, but the question that didn’t quite surface in the room is probably the one that matters most:

Who is governing the AI that people are already using, and what happens to the personal information contained in it? Because people are using it, for decisions, in moments that aren’t hypothetical, and most disturbingly in moments that aren’t psychologically stable.

The systems are already in the building. The governance isn’t.

I enjoyed the event, and admire the work that Mark and Michelle are doing. It is thoughtful, necessary and timely, but the gap between principle and practice is where this gets interesting.

AI in mental health. AI in HR. AI anywhere near human vulnerability may not be as benign as it looks. Not because the technology is inherently harmful, but because it’s being used generally in ways that haven’t been thought through.

Turing asked whether a machine could think. Seventy years later, the vast majority of people, me included, are using those machines to listen, process and create. They’re being told things that humans don’t feel they can talk to other humans about. Whether that be on a personal or business level, and they are being told things they should not be, shared things they should not be and that data, often highly personal, highly sensitive data is being reviewed, analysed and processed to feedback into the ecosystem, creating a feed loop.

Now being the cynic I am, I don’t trust big tech and their motives. That’s not based on feelings, it’s documented fact. Data, your data, your personal, genetic, health and emotional data is the new gold.

A $340 billion global industry built on collecting, analysing, and selling personal data, including emotional, location, habits, preferences, biometrics the list goes on [9].

In the 1849 California goldrush, 300,000 people rushed to search for gold, about half made a modest profit, most left with roughly what they came with. The people who got rich were the ones selling shovels. So if digging for data is the modern day goldrush who’s selling the shovels, and what are they doing with what the shovel digs up?

The Birss ruling touches a raw nerve, but if nearly half of employees are using personal AI accounts at work how much employment data including salary, performance and grievance data has already been compromised without anyone knowing?

How much of that is already being sold, to who, by whom and what is happening to it?

Thoughts please….

Sam

Configure YOUR system. contAIn the chaos. Control YOUR outcome.

Sources

[1] Turing, A.M. (1950). “Computing Machinery and Intelligence.” Mind, 59(236), 433-460.

[2] McCarthy, J., Minsky, M.L., Rochester, N., & Shannon, C.E. (1955). “A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence.”

[3] Ongoing litigation involving AI and mental health harms (various filings, 2025–2026).

[4] Reported case filings involving psychiatric hospitalisation linked to AI interaction.

[5] Social Media Victims Law Center filings alleging “suicide coach” behaviour.

[6] UK v Secretary of State for the Home Department [2026] UKUT 81 (IAC); remarks by Sir Colin Birss, Chancellor of the High Court.

[7] Netskope (2026). Cloud and Threat Report: AI in the Workplace.

[8] Lenovo (2026). Work Reborn Research Series: The Shadow AI Challenge.

[9] Mordor Intelligence (2026). Data Broker Market Size, Growth, Trends & Forecast Report 2031. Global data broker market valued at $294.27 billion in 2025, projected to reach $448.32 billion by 2031.

This article was originally published on Medium. Full sources and references are available there.