A European train wreck with AI in the driver’s seat

Welcome to my home country! No, it's not Kazakhstan (Jak si? masz, Borat!), but it is beginning to look like it. I am from the Netherlands, where government incompetence is the system.

And in my country, there has been a huge wrong doing, caused by the tax department and by AI.

The “Toeslagenaffaire” as it is called (don't try to pronounce: the "social benefits scandal") is a perfect example of how a government managed to turn a welfare program into a human rights violation.

And in the middle of this disaster is an AI algorithm that was allowed to run wild, unregulated, and unchecked, like a toddler in a candy store.

Only this time, people lost their homes, their children, and in some tragic cases, their lives.

And if you think that the Dutch Government has learned from this, you are mistaken. The Public Employment Service UWV has another trick up its sleave.....

Support us and sign the petition that we have started to stop the government in using illegal tactics to train AI, click here...

Before we start!

If you like this topic and you want to support me:

1. Comment on the article; LinkedIn appreciates that and it will really help spread the word ??
2. Connect with me on Linkedin ??
3. Subscribe to TechTonic Shifts to get your daily dose of tech ??
4. TechTonic Shifts has a new blog! With lots of other goodies besides text. ??
5. Sign the petition to stop the NL Government from using our personal data to train AI ??

Social benefits was a lifeline that became a noose

Social benefits are invented to help people.

They were designed to give lower-income families financial support.

For instance, childcare allowances helped parents afford daycare, so they could work and contribute to society.

It was a beautiful concept.

And it still could have been, if tax people would have taken responsibility.

But like most beautiful concepts in the hands of the government, it was turned into a bureaucratic weapon.

Families that needed help found themselves on the receiving end of letters telling them they owed the government tens of thousands to hundreds of thousands of euros.

Why do you think that was the case?

Because they were considered to be “fraudsters.”

But their only crime was needing assistance.

Oh, and having a last name that didn’t sound Dutch enough (yet not in every case though).

But more on that later.

AI Profiling was a weapon of mass destruction

How did these families get labeled as fraudsters?

My beloved AI.

The Dutch tax department (Belastingdienst, just try to pronounce it), in all its wisdom, decided to use an algorithm to identify fraud.

The algorithm did not have any humans in the loop.

Now, AI can be great, when it’s used properly. But this AI wasn’t trained for ethics or fairness. It was trained to do one thing: profile.

It did not care about due process, or context, or basic human decency.

The algorithm scanned through the data and flagged families left and right. And they were often based on ethnic background and minor administrative errors. But not all the times. It also flagged random people as criminals.

No human oversight. No accountability. Just cold, hard code.

The AI flagged anyone it “thought” looked suspicious.

It just kept going in circles, pulling more and more innocent people into its orbit of destruction. A bit like a dog chasing its tail

And the government let it run wild for years.

Years!

As long as the algorithm said it was working, the tax department collected the money, who cared about the human lives being torn apart

Lives in ruins, people killed even.

But hey, the Algorithm did its thang!

Let’s talk about the consequences of this reckless use of AI.

Thousands of families were labeled as fraudsters. They were forced to pay back absurd amounts of money.

Working class families who could barely make ends meet, because the NL has one of the highest and harshest tax regimes in Europe (and even the world), were suddenly slapped with bills they had no hope of paying.

People lost their homes. Some were forced to give up their children to foster care because they couldn’t provide for them. And then there’s the tragic reality: some victims were driven to suicide because they lost everything, from their homes, their jobs, and their kids.

I personally know two people that had to go through this ordeal. One is still alive and is fighting back. The other one has left two kids in destitude.

The shame, the stress, the financial ruin, it was too much to bear.

And that was all because of an algorithm with no moral compass and a tax department and a government that couldn’t be bothered to care as long as it brought in the dough.

Writing about this major injustice, brings me to tears alone...

No kidding this time.

If you want to jump to the petition right away, click here...

Racial profiling in action

The AI was xenophobic.

A report by Amnesty International titled "Xenophobic Machines" exposed how racial and ethnic discrimination was baked into the design.

The system flagged people of non-Dutch origin more often for further scrutiny.

Having a second nationality made you a prime target for this data-driven witch hunt.

The AI not only reinforced but amplified existing biases.

Instead of treating applicants equally, it created a discriminatory loop, feeding on its own flawed data to flag more non-Dutch nationals.

This was a human rights violation to the max.

You can read more about this in Amnesty International's scathing analysis of the scandal, which rightly pointed out that the Dutch government, in its rush to automate public services, has left the most marginalized in society to pay the highest price.

Literally.

The system was so opaque that when a person was flagged, the civil servant reviewing the case had no idea why.

No explanation, just a red flag from the almighty AI. This "black box" approach to governance is as far from transparency as you can get.

No Human in the Loop caused the perfect storm

What happens when you let a machine run wild with no one watching the controls?

In one word: Disaster.

No Explainability, and as it turned out, in the end no Accountability as well...

Forget it.

The AI was designed to be inscrutable, leaving no one accountable for the decisions it made.

This lack of transparency was one of the primary reasons the whole system went off the rails.

Without human oversight, AI becomes a ticking time bomb, as it did in this case.

The lone crusader in a sea of incompetence

Enter Pieter Omtzigt, the man, the myth, the hero.

While everyone else in the Dutch political machine was content to look the other way, Omtzigt took up the fight. For years, he battled the system, calling out the injustice, demanding answers, and refusing to let the victims be forgotten.

Of course, his colleagues didn’t appreciate his relentless pursuit of truth and justice. During an election year, they tried to push him aside with the now infamous “functie elders” suggestion (essentially a “give him another job), so he’ll shut up” move.

But Omtzigt didn’t shut up.

A note leaked to the press, and the entire fiasco backfired.

Huge.

Suddenly, the public was paying attention.

But even then, justice didn’t come swiftly.

Sure, a few people got fired, but the wrong ones.

The people who were complicit are still in their cushy tax dept. and government jobs. Still pulling the levers of a broken system. Omtzigt may have won the battle, but the war for compensation and accountability is still dragging on.

A legal obligation the Tax Department conveniently ignored

And here is where it starts to get out of control...

Even the courts ruled that the victims had to be compensated for the incorrect profiling and their loss of... everything.

The law says the tax department has to make things right. But this is the Tax Department (with an incompetent capital T and D) we’re talking about, they don’t care much for laws that inconvenience them.

So, instead of compensating people, they’ve been dragging their feet.

It has been years since the ruling, and many victims are still waiting for justice. The tax department, in its infinite wisdom, has decided to move at a snail’s pace. Maybe they’re hoping people will just give up and go away.

After all, it worked for them before.

But who is on Algorithms overwatch?

You would think that the Dutch government would have learned from the benefits debacle and tightened up their AI governance.

But you’d be wrong.

The Ministry of Economic Affairs (EZ) is responsible for registering all the algorithms used by the government, and the Privacy Authority is responsible for upholding the law. In theory, they’re supposed to make sure these algorithms are ethical, transparent, and accountable.

In reality? They don’t even know how many algorithms are out there.

They are giving a kid a box of matches and telling them to “play responsibly.”

There’s no oversight, no regulation, and no clue what’s going on behind the scenes.

Just a EU Act paper tiger

The Dutch Privacy Authority is MIA when It comes to AI

Ah, the Privacy Authority (Autoriteit Persoonsgegevens).

That is the agency that’s supposed to protect the privacy of Dutch citizens. The watchdog that’s supposed to ensure AI systems aren’t violating people’s rights.

But when it comes to AI they are about as useful as a chocolate teapot.

They don’t know how to handle AI.

They don’t understand the technology, and the ethics behind it. No vision at all.

And when it comes to enforcing privacy laws around AI, they’ve been missing in action.

Want another example?

Let’s talk about the Dutch public employment service, UWV.

Another one: UWV’s €65 Million AI debacle

In its infinite wisdom, the UWV spent €65 million on an AI matching tool for their public employment service.

That’s right, €65 million.

For a tool that used illegally obtained data to train its algorithm.

Only because it has the word "AI" in it...

I wish I were making this up.

And their previous system only cost them a factor 10 less.

Not only was the algorithm built on shaky ethical grounds, but it didn’t even work properly.

People were mismatched with jobs, data privacy laws were violated, and the whole thing is going to be a disaster from start to finish.

Did the Dutch government learn from it's previous mess?

Of course not.

Algorithms that deal with people's lives are classified as high risk Algorithms and should have a plan about accountability, transparency of the algorithm, risks accociated with it, privacy... But there is nothing you can find about it ...

And the Privacy Authority is knowhere to be found this time as well.

A new disaster in the making.......

If you want to jump to the petition right away, click here...

Meet Marc Drees. The Don Quixote of AI transparency

Thankfully, not everyone is content to let the government get away with these AI crimes.

Enter Marc Drees, the man behind RecruitmentMatters (only in Dutch I'm sorry folks)

Marc Drees , is a far relative of one of Netherlands' most famous prime ministers, and a true advocate for justice. He is a tireless advocate for transparency in government-run AI systems.

He has been calling out the UWV, the Privacy Authority, and anyone else who thinks they can use AI without accountability. Heck, even LinkedIn for allowing the public employment service to scrape our data in the first place.

Well, it takes one to know one... read: LinkedIn siphons user information to Train AI, skips disclosure

Like Omtzigt, Drees has been battling the bureaucratic machine, often with little support.

But he’s not backing down, and neither should we.

Help us stop UWV from using illegally obtained data

Stop UWV from using illegally obtained data for training

We demand that the Dutch Government, and the Privacy Authority AP, let's UWV immediately stop using illegally sourced data and adopt transparent, ethical AI practices that respect the privacy of Dutch citizens.

Sign this petition to protect our rights and ensure accountability in public services.

HCAI is the Hero that we desperately need

So, what’s the solution to this AI nightmare...

Human-Centric AI is a necessity.

HCAI emphasizes transparency, ethics, accountability, and, most importantly, human oversight. AI can’t be left to run wild because it needs guardrails. It needs people in the loop who understand ethics, law, and basic human decency.

HCAI is about making sure AI systems serve people, not destroy them.

It’s about preventing benefits-style disasters by making sure that AI decisions are transparent and fair. And it’s about integrating these principles with privacy laws so that we can finally bring some sanity to the way the government uses technology.

Why HCAI must be the future of AI governance

The benefit scandal and the UWV debacle are early warnings. Big warnings.

They show us what happens when we let AI systems run without oversight, without ethics, and without accountability.

And to me, HCAI offers a path forward, a way to prevent these disasters from happening again.

But for HCAI to work, it needs to be integrated with privacy laws.

AI systems must be transparent, accountable, and subject to the same laws as any other government tool. The days of letting algorithms make life-altering decisions without human intervention must end.

AI is a Tool, and not a WMD, unless you let the Dutch government use it

At the end of the day, AI is just a tool.

But in the hands of the Dutch government, it became a weapon of mass destruction.

Whether it’s the tax department’s fraud-hunting algorithm or the UWV’s illegal data-driven job-matching tool, the government has proven time and time again that they cannot be trusted with AI.

It’s time for real reform.

It’s time for HCAI.

And it’s time for the government to take responsibility for the lives they’ve destroyed. Because if they don’t, the next AI disaster is just around the corner.

Read more about this AI scandal:

1. Victims of Dutch child-care benefits scandal struggle to move on
2. Victims of childcare benefit scandal press minister for faster compensation
3. Review

Well, that's a wrap for today. Tomorrow, I'll have a fresh episode of TechTonic Shifts for you. If you enjoy my writing and want to support my work, feel free to buy me a coffee ??

Think a friend would enjoy this too? Share the newsletter and let them join the conversation. LinkedIn appreciates your likes by making my articles available to more readers.

Signing off - Marco

Top-rated articles:

• Human-Centered AI design and the HAX toolkit | LinkedIn
• The next big thing is Tiny (ML)! | LinkedIn
• A model for strategy development in the AI age | LinkedIn
• How AI is preventing cyber attacks [including e-book] | LinkedIn
• The Machine Learning Lifecycle and MLOps | LinkedIn
• 150 Digital Transformation templates new ones added | LinkedIn
• Knowledge graphs for Machine Learning are so cool ! | LinkedIn
• The AI Readiness Index - a brilliant framework for assessing AI adoption in your organization | LinkedIn
• What is explainable AI (XAI) | LinkedIn