The High-Risk HR AI Delay: Why the EU AI Act Still Matters

Hello Humans And welcome to the podcast that aims to summarise each week an important AI development in five minutes or less. My name is Barry Phillips. People think that because I talk a lot about AI, I must be positively evangelical about it all the time. You know the sort of thing: every development is a good one and every new tool is progress.

In fact, I’m not. Yesterday, I renewed my house insurance, and AI, or at least automation, probably cost me the best part of £100. Let me explain. The end of our garden is also a river edge.

Lovely during these sunny days, but not ideal if you are looking for bargain house insurance. There’s a box on the insurance form that asks: “Is your house located within 50 metres of a river, lake or sea?” Of course, I have to tick yes. But there’s no accompanying box where I can explain that there’s never been a flood in the 15 years I’ve lived in the house.

No little space saying, “Please insert useful human context here.” No opportunity to say, “Yes, there’s a river, but no, I am not currently paddling through the kitchen.” The system sees one data point. River nearby.

Risk goes up. Price goes up. It’s now down to me to well take it or leave it. And this is the bit that matters for HR.

Where else could AI, or automated decision-making, be working against your interests without you really understanding it? Recruitment? Promotion? Performance management?

Absence monitoring? Shift allocation? Redundancy scoring? Training opportunities?

These are not small matters. These are decisions that affect livelihoods, confidence, reputation, careers and, frankly, whether someone feels they are being treated like a person or a spreadsheet stat. So why isn’t there a law to protect individuals against this sort of thing? Well, there is.

It’s called the EU AI Act. Now, before we get carried away, the AI Act is not a magic wand. It will not refund my £100. It will not make every awkward insurance portal suddenly develop empathy.

And, being legally picky for a moment, not all automation is AI, and not every frustrating automated decision will fall neatly under the Act. But the direction of travel is important. The EU AI Act says, in effect, that when AI is used in areas that can seriously affect people’s lives, we need more than blind faith and a privacy notice nobody reads. We need risk management.

We need transparency. We need human oversight. We need proper records. We need accountability.

And in HR, that really matters. Because AI in the workplace can be brilliant. It can help write job adverts, summarise policies, support learning, spot patterns, reduce admin and give HR teams back time they badly need. Used well, it can make HR more human, not less.

But used badly, it can quietly bake unfairness into decisions at scale. It can filter out candidates who never know why they were rejected. It can flag employees as “low performers” based on dodgy data. It can reward the loud, the visible and the easily measurable, while missing the careful, the quiet and the brilliant.

That is why the EU AI Act is such a significant achievement. Whatever your politics, stand back for a second and admire the scale of it. A political union of 27 member states, working across 24 official languages, has tried to create a single legal architecture for one of the fastest-moving technologies in human history. And it’s protecting well over 300 million people.

The Act has to deal with banned practices, high-risk systems, transparency duties, general-purpose AI, enforcement, regulators, innovation, SMEs and cross-border markets. For HR, the message is clear. Don’t wait until the law forces your hand. Start asking better questions now.

What AI are we using? Who is affected? What data is it using? Can we explain the result?

Can a human challenge it? Would we be comfortable defending this decision to the employee, a tribunal, a regulator or, worst of all, our own conscience? Because the future of AI in HR will not be decided by the cleverness of the tools. It will be decided by the quality of the judgement around them.

AI can process. AI can predict. AI can recommend. But it cannot care.

That bit is still ours. And if HR gives that away, we won’t just have automated a few tasks. We’ll have automated away the very thing that makes HR worth having in the first place. Until next week, bye for now!