Legislating Against AI Job Displacement – Do we need an AI tax?
…with a bonus Tony Stark lesson for Humanity!
Displacement is already underway
Artificial Intelligence is no longer a distant future potential disruption coming onto our horizon. In “AI’s impact on Europe’s job market: A call for a Social Compact”, McKinsey reported that the share of firms using AI in at least one business function rose from 20% in 2017 to 78% in 2024, driven largely by the explosion in generative AI tools, with adoption of generative AI alone surging from 33% to 71% between 2023 and 2024.
So it is clear that AI is rising fast and yet despite the memes that fly around inside the AI Hype-bubble about “AI will make our lives better”, many firms have been letting staff go in their thousands recently while at the same time investing millions in AI programmes and additional GPU capacity. Oracle, Meta and Block have all had big staff cuts and blamed them on AI while Venture Capital firms, who have invested heavily in AI firms over the last number of years are pushing harder and harder to make the business case of Enterprise AI take-up. Unfortunately, this “Business Case” often only really works if workers carry the actual cost by losing their jobs so that AI can claim the savings to make the return on investment work.
Regardless of whether we believe that the layoffs are truly being caused by AI or not, the stakes for workers couldn’t be higher. In “Artificial Intelligence impact on the Global Job Market (2025–2030)”, by the Congruence Foundation Research Team, Goldman Sachs projected 300 million jobs “at risk”, while the World Economic Forum estimates AI will displace approximately 92 million jobs. Incidentally they also say that 170 million new roles could be created by 2030 but even this potential net gain masks enormous structural disruption to the people in the jobs that exist today. In the United States alone, 55,000 jobs were impacted by supposed AI-driven automation in 2025 but in the first quarter of 2026, roughly 80,000 US tech jobs were cut, with nearly half attributed directly to AI.
But before I get to talking about the purpose of this article which is countries pushing back using legislation, economic frameworks, or regulatory instruments, I first want to talk about what this anticipated unemployment will actually do to the human beings that will be affected. Because without that, this would just be a policy conversation and for the people who know or work with me they know very well I don’t do policy conversations.
What unemployment does to people
I grew up in Ireland in the 1970s and 80s and if I tell you that it was one of the most depressing places on earth at that time it would not be an exaggeration mainly because of one problem: unemployment. Between 1971 and 1988, unemployment in Ireland rose from 5.5% to 18%. That meant that roughly 1 person in every 5 was long-term unemployed. By 1989, 45,000 people were emigrating per year mainly to the UK, US and even farther afield. I was luckily in college during this period but all around me I was seeing people losing their jobs literally on a daily basis. I witnessed many of my own neighbours losing their jobs in their 40s or early 50s and some of them simply never worked again for the rest of their lives. Eventually in the 1990s the situation improved mainly because of an innovative corporation tax initiative, which still exists to this day, to attract Foreign Direct Investment and create badly needed jobs for the population.
For the last 30 years or so, the world, or more correctly the Global West, has enjoyed prosperity on a level where many countries have achieved almost full employment. For the generations born during this time, they have not experienced unemployment and so are very unprepared for the inevitability of what is coming if nothing is done.
When I hear Technology CEOs like Altman, Musk or Zuckerberg saying things like “In the future, work will be optional” or “Governments can simply hand out free social money”, I just shake my head both in disappointment and in despair. These “Tech Bros” have no idea what unemployment feels like. They have never experienced it in their lives and definitely never will, hidden away as they are in their Tech-Billionaire bubbles.
“Unemployment” is a sterile sanitised word that sounds just like a spreadsheet category. It absolutely is not just that. It is a scourge that destroys people and societies. It is one of the most devastating things that can happen to a person in their lifetime and the data is clear and unambiguous to support the following direct effects of unemployment.
It actually kills people!
“Financial Stress, Unemployment, and Suicide – A Meta-Analysis, by the McGill University in Montreal, found that unemployment is associated with a 58% higher relative risk of suicide compared to the employed population. The suicide attempt rate among young unemployed people is three times higher than among young working people. While both genders are affected, the risk is particularly high for young men. The risk peaks in the first five years of unemployment and persists at an elevated level for up to 16 years (Yes you read it correctly 16 years!) after job loss. The longer the duration of unemployment, the higher the frequency of suicidal thoughts and attempts. Critically, regaining employment led to a measurable and significant reduction in suicidal thoughts. That is not a coincidence. Like my headline said unemployment actually kills people.
It destroys mental health at scale
Various research studies have consistently linked unemployment to on average a 39% increased risk for mood disorders and a 21% increased risk for anxiety disorders including a collapse in the feeling of self-worth. Every 1% increase in global unemployment rates is associated with a 1% increase in male deaths by suicide. Long-term unemployment, i.e. beyond 52 weeks correlates with large, lasting negative effects on a person’s mental health, often independent of any pre-existing condition.
It drives violence in the home
Research using administrative data from Brazil covering 2 million domestic violence cases across a decade found that job loss has a significant causal effect on domestic violence, with increases of over 30%. Financial hardship and unemployment have been confirmed as strong contributors to intimate partner violence. Economic downturns in countries are also often correlated with increased calls to domestic violence hotlines. Children grow up in these violent households and the damage compounds across generations.
It destroys physical health
Chronic stress, which unemployment always produces reliably and relentlessly, is a predominant risk factor for a range of neuropsychiatric and physical health conditions including depression, anxiety, cardiovascular disease, and immune dysfunction. Between 1933 and 1939 in the US, President Franklin D. Roosevelt introduced his “New Deal”, aimed at providing relief, recovery, and reform to combat the effects of the Great Depression. Many years later studies have shown that “New Deal” relief spending in the United States estimated that every additional $153,000 in relief spending (in 1935 dollars) was associated with a reduction of one infant death, one suicide, and 2.4 deaths from infectious diseases. Relief spending, in this particular case giving people Work saved lives measurably and directly. I’ll come back to this in the next section.
It steals people’s dignity
This is harder to put in a study but easier to recognise in a person. Work is not just about income. It is about identity, structure, social belonging and self-worth. When it is taken away, especially when it is taken away by an algorithm that nobody voted for, the psychological damage is profound and immediate. As WPA Administrator Harry Hopkins observed when designing Roosevelt’s New Deal, “subjecting workers to means tests to prove their destitution fostered the wholesale degradation of their finest sensibilities”. He insisted on paid work, not handouts, for precisely this reason.
All of these effects are not edge-cases, they are all predictable, well-documented, population-level outcomes and they will all happen if AI job displacement is allowed to happen on the scale that is predicted.
What history tells us about mass-unemployment and what actually worked
Humanity has faced catastrophic unemployment before, but Humanity does not have to take this lying down. The purpose of any Government is to protect the rights of the most vulnerable people in its society and right now, many hundreds of thousands of workers and college graduates are in a vulnerable position. They are the reason that legislation protecting people from AI-driven job displacement is not an economic luxury, it is a moral obligation for every Government. Throughout history when faced with mass unemployment events the response that always preserved both economies and human beings has consistently been the same: Governments intervening directly to create, protect, and dignify human work. The following are some examples of this in action.
The United States New Deal (1933–1943)
When Franklin Roosevelt took office in 1933, 25% of the American workforce was unemployed. The banking system had collapsed. Hunger was widespread. Roosevelt’s response was the “New Deal” which consisted of a suite of the most ambitious public works programmes in history.
The Works Progress Administration (WPA), established in 1935 with a first appropriation of $4.9 billion (approximately 6.7% of 1935 GDP) supplied paid jobs to the unemployed while building public infrastructure. At its peak in 1938, it employed more than 3.3 million Americans. Between 1935 and 1943, it employed 8.5 million people in total. These were not made-up-work jobs. The WPA built more than 4,000 new school buildings, 130 new hospitals, 29,000 new bridges, 150 new airfields, and paved or repaired 280,000 miles of roads. It planted 24 million trees. It employed musicians, artists, writers, actors and directors through Federal Project Number One, because the Roosevelt administration understood very well that the soul also needs work just as much as the body does.
The Civilian Conservation Corps (CCC), running from 1933 to 1942, employed young unemployed men aged 18 to 25 on conservation and park-building projects. At its peak it enrolled more than 500,000 men. Modern research using linked government records shows that longer CCC service significantly improved employees’ long-term health, nutrition, and lifetime earnings. It did not just give people jobs in the short term. It changed their life trajectories for the better even into later generations.
The core philosophy, articulated by Roosevelt himself, was this: “Providing useful work is superior to any and every kind of dole.” The distinction mattered. Direct relief or in other words receiving money for nothing, eroded the dignity the programmes were designed to restore. Only humans working to earn their living preserved it. Now you know why my earlier reference to the flippancy shown by the Tech billionaires in this regard is not only ignorant of history but it shows a contempt for the human condition.
Germany’s Public Works Programmes (1932–1936)
Germany in the early 1930s, after the Wall Street crash, faced unemployment of close to 30%. 6 million people were out of work by 1933. The public works response, initially planned under the Weimar Republic and dramatically expanded after 1933, included large-scale infrastructure construction, most famously the Autobahn highway network, which at its peak employed 125,000 men directly, with hundreds of thousands more working indirectly through supply chains etc. Other projects included hospital construction, housing, railway expansion, and land reclamation. Official figures showed the programmes reduced unemployment by more than four million people within a few years.
Of course we all know that the political context of this era in Germany was catastrophic and the public works programmes ended up being intertwined with forced labour, rearmament, and the violent suppression of all worker rights. The lesson here is emphatically not to ever dignify the politics, but to recognise that mass unemployment creates conditions of desperation that authoritarian movements have historically exploited and that providing dignified work is therefore not only a humanitarian imperative, but a democratic one as well.
India’s MGNREGA (2005–Present)
The Mahatma Gandhi National Rural Employment Guarantee Act is one of the world’s largest public employment schemes, guaranteeing up to 100 days of paid manual work per year to any rural household whose adult members are willing to do unskilled labour. At its peak, it has employed over 50 million households annually. Importantly, it is demand-driven i.e. work is provided when people request it, not when the government decides to offer it. Independent research has found it reduces rural poverty, increases rural wages through competitive pressure, and improves food security. It is imperfect and subject to implementation challenges in parts, but it represents a model of using public employment as a genuine social safety net, not as charity, but as a basic human right to work.
The pattern across all of these is consistent: when technology or economic shock threatens mass unemployment, governments that respond with creating or protecting dignified, purposeful human work create better outcomes socially, economically, and in terms of human health than those that respond with passive benefits or nothing at all.
What the world is doing now about AI
Against this backdrop, governments around the world are finally beginning to legislate. Here is what has actually been enacted or passed. Not promised, not proposed, but actually done.
🇪🇺 The European Union — The AI Act (2024)
The EU’s average unemployment rate sits around 6%, but the threat ahead is severe. Researchers have warned that generative AI could eliminate up to half of all entry-level white-collar jobs and push unemployment up by up to 20 percent within just one to five years.
In 2024, the European Union Council made history by approving the EU AI Act, the first-of-its-kind comprehensive law for artificial intelligence, designed to mitigate the risks and challenges of AI systems across all 27 EU member states. As a general guiding principle, “entities must not use AI in ways that endanger safety, rights, and livelihoods”.
Under the Act, AI systems intended to be used for recruitment or selection — including those that place targeted job advertisements, analyse and filter applications, and evaluate candidates are classified as “high-risk” systems, subject to the strictest obligations. Employers’ use of AI in the workplace will be treated as potentially “high risk,” triggering duties such as worker notice, human-in-the-loop oversight, monitoring for discrimination, and logging. From February 2, 2025 onward, emotion recognition in workplaces was banned outright.
Critics note that while the Act addresses transparency and bias, it ultimately overlooks the critical issue of employment security which many see as the main cause behind workers’ distrust of AI so overall it’s a good start but it’s not quite there yet.
🇮🇹 Italy Law No. 132/2025 (The First National AI Law in the EU)
Italy’s unemployment rate fell to 6.8% in May 2024, still above the OECD average of 4.9%, while Italy’s employment rate remains well below the OECD average. More than 22% of those aged 15 to 24 were without a job in 2023 and Italy remains among the EU member states with the largest share of unemployed people, recording the worst unemployment figures among G7 countries.
On September 17, 2025, the Italian Parliament approved the first national AI law in the European Union, Law No. 132/2025, making Italy the first EU country to adopt domestic legislation governing the use of AI, effective October 10, 2025. The employment provisions are explicit and among the strongest in the world including:
Prohibition of fully automated employment decisions. All employment-related decisions must involve meaningful human oversight and complete automation is not permitted.
Employers are required to consult with trade unions on the use of AI systems in the workplace.
Workers must be informed of the criteria and logic behind any automated decisions that affect them, with safeguards to prevent discrimination and unjust treatment.
It also establishes a National Observatory to monitor the impact of AI on the labour market, promote training, identify sectors and professions most affected, and propose practical solutions to manage these changes.
Italy has drawn a line in the sand. A machine cannot fire you, demote you, or pass you over for promotion without a human being accountable for that decision.
🇰🇷 South Korea The AI Basic Act (2025/2026)
South Korea’s headline unemployment rate is low at around 3%, but the structural picture is more complex. Professional occupations in Korea will likely benefit from AI, while clerical jobs are at risk of displacement and Korea has a significant portion of its workforce employed in clerical roles. These jobs face replacement rather than augmentation.
South Korea became the second country in the world, following the EU, to enact a comprehensive regulatory law on artificial intelligence. The Framework Act on the Development of Artificial Intelligence and Establishment of Trust, which integrated 19 separate bills, passed the National Assembly on December 26, 2024, with overwhelming bipartisan support and took effect January 22, 2026. The Act requires clear explanations of the key criteria behind AI’s final results, while mandating the government to develop policies to help residents adapt to AI-driven societal changes and that AI must be designed to avoid unjust bias, particularly in employment.
🇧🇷 Brazil — AI Bill No. 2338/2023 (In Progress)
Brazil’s formal unemployment rate was around 6.2% at end of 2024, but its gig-economy tells a harder story. The growing use of AI and algorithms in managing labour relations, particularly in platform economy companies like Uber, iFood, and Rappi has sparked intense legal debate, with labour management carried out almost entirely through algorithms that define routes, evaluate performance, assign tasks, and can disconnect workers with no human involved in the decision.
On December 10, 2024, the Brazilian Senate approved Bill No. 2338/2023. Unlike the EU AI Act, Brazil’s legislation includes specific obligations to mitigate negative impacts on workers such as job displacement and to promote continuous training and skill-building programmes. The bill explicitly mandates worker-protection against algorithmic management and requires transparency in automated decision-making affecting employment. It is now under review by the Chamber of Deputies.
🇺🇸 The United States State Laws
The US has no federal AI employment law, but the states are moving fast. In 2024 alone, over 400 AI-related bills were introduced across 41 US states. Key enacted laws include:
New York City: An anti-AI bias law in effect since 2023 requires employers to conduct annual independent bias audits of automated employment decision tools used in hiring and promotion.
Colorado: The Colorado Artificial Intelligence Act, effective February 1, 2026, protects against algorithmic discrimination in “consequential decisions” concerning employees.
Illinois: Effective since January 1, 2026, requires all employers to notify employees and applicants when AI is used to make employment decisions.
California: Regulations effective October 1, 2025 state that any automated decision system used in employment must have meaningful human oversight, with someone trained and empowered to override the AI.
🇨🇳 China Courts Rule It Illegal to Fire Workers Because of AI (April 2026)
Saving the best until last, this one only landed last week and it is the most direct response to AI displacement anywhere in the world!
China’s urban youth unemployment rate reached 15.3% in March 2026, in an economy already facing deflation, a property crisis, and weak consumer demand. Meanwhile, China’s core AI industry exceeded 1.2 trillion yuan in 2025. The political tension between those two facts, a booming AI sector and a struggling youth labour market, sits directly behind these rulings.
In two separate court rulings, one from Hangzhou and one from Beijing, Chinese courts have established that companies “cannot” fire workers simply to replace them with AI. This is now a completely illegal practice and will be punishable on the company.
The Hangzhou case involved a quality assurance supervisor identified only as Zhou, who joined a technology company in November 2022 and earned 25,000 yuan (approximately $3,640) per month reviewing and filtering AI outputs. In 2024, the company decided its AI systems had improved to the point where Zhou’s role could be automated. It offered him a demotion to a role paying 15,000 yuan, a 40% pay cut. He refused and the company fired him. Zhou won at arbitration. The company appealed but he won at the lower court. The company appealed again but he won again at the intermediate court. Three levels of adjudication. Same result each time.
The legal reasoning is as precise as it is significant. China’s Labour Contract Law permits termination when “objective circumstances materially change” and render a contract unperformable, a standard typically reserved for events like natural disasters, company relocations, or policy changes that destroy a business. The company argued that AI adoption met this standard. The court rejected this argument in unusually direct language: adopting AI was a “voluntary move to stay competitive”, a deliberate, foreseeable business decision made by the employer. By characterising it as an unforeseeable change, the court found, the company had “shifted the risks” of its own strategic choices onto its employee and that is simply not permitted.
The Beijing case, published by the Beijing Municipal Human Resources and Social Security Bureau in December 2025 as one of its ten most significant labour arbitration decisions of the year, involved a map data collector named Liu, hired in 2009. When the firm switched to AI-based data collection in 2024 and eliminated his division, Liu was dismissed. He also won arbitration. The bureau ruled that the AI pivot was a deliberate, predictable strategy, not unforeseeable, and that firing Liu illegally shifted the business risk to the employee.
The Hangzhou court published Zhou’s case in a bundle of “typical examples” of AI labour disputes, timed deliberately for International Workers’ Day on May 1, 2026. That timing was of course not accidental!
China’s courts have not said companies cannot use AI. They have said companies cannot use AI as a pretext to fire people. The distinction forces you to think if you automate a role, you must find another role for the person who held it, at comparable terms or prove genuine grounds for redundancy entirely unrelated to the automation itself. That is expensive. It is also, the courts have decided, the law. As one legal commentator observed bluntly: “The country we keep calling authoritarian just gave its workers more legal protection from AI than any democracy on Earth.”
The United States, by contrast, has no equivalent protection. American employment law operates on an at-will basis in every state except Montana, meaning employers can terminate workers for any reason not specifically prohibited by statute and being replaced by AI is not a prohibited reason.
A new idea worth taking seriously: The AI Tax
The legislative approaches described above, transparency requirements, bias audits, consultation mandates, court-enforced reinstatement rights are all meaningful. But they share a structural limitation: they make AI-driven dismissals harder or slower, without making them financially painful at the scale that actually changes a board’s calculus.
I want to draw attention to a suggestion to solve this problem which is based on a parallel model that already exists, that already works, and that could be applied directly to this problem quickly. It has been operating quietly across Europe and Asia for over a century and it is called the disability employment quota-levy system.
How the disability levy system works today
In Germany, all employers with 20 or more employees are legally required to ensure that at least 5% of their workforce consists of people with disabilities. If they fail to meet that quota, they pay a compensatory levy for every unfilled position currently between €140 and €720 per month per vacancy, scaling upward based on how far below the quota the employer falls. If no disabled person is employed at all, the maximum levy applies.
In France, the threshold is 6% of the workforce for companies with 20 or more employees. Employers who fail to meet the quota contribute to a national fund (AGEFIPH) at a rate equivalent to 400 to 600 times the gross hourly minimum wage per unfilled position and that amount triples for employers who have failed to meet the quota for three consecutive years.
In Italy, Japan, Spain, and dozens of other countries, similar systems operate. The International Labour Organisation has explicitly noted that “the main purpose of the levy is to encourage employers to meet their quota target, not to raise revenue.” The levy is a behaviour-change mechanism. It makes the financially convenient choice of ignoring a social obligation, more expensive than the socially responsible choice.
Crucially, the German government has been explicit that payment of the levy is not an admissible substitute for the employment obligation. You cannot simply buy your way out of it. The obligation remains, and the levy exists to accelerate compliance.
My proposal: An AI Displacement Levy (The AI Tax)
The same logic applies, with an obvious and compelling directness, to AI-driven workforce reduction. Here is how a workable system could be structured:
The trigger: Any company above a defined size threshold say, 250 employees, that reduces its human headcount in a given function by more than a specified percentage (for example, 10% over 12 months) in circumstances where AI adoption has been documented in that function, becomes subject to the levy.
The calculation: The levy is assessed per displaced worker, per quarter, at a rate proportionate to the company’s AI capital expenditure. A company pouring $50 million into AI infrastructure while simultaneously cutting 500 jobs is making a specific financial trade-off. The levy makes that trade-off more expensive. A straightforward formula might be something like: Levy = (number of displaced workers) × (median annual wage in the relevant sector) × (levy rate), where the levy rate scales with the ratio of AI investment to workforce reduction.
The fund: All proceeds go directly into a ring-fenced National AI Transition Fund, used exclusively for: retraining and upskilling programmes for displaced workers; publicly funded job guarantee schemes in infrastructure, green energy, healthcare, and other areas of national need; income bridging support for workers in transition; and independent research into AI’s actual employment impact.
The exemption: Companies that retrain and redeploy displaced workers into comparable roles rather than dismissing them pay no levy. This mirrors the disability quota model exactly where the goal is not revenue extraction, it is behaviour change. Boards that redirect AI efficiency gains into workforce transformation rather than headcount reduction are rewarded. Boards that pocket the savings while externalising the human cost are penalised.
Why this model works where others don’t
The reason the disability levy has endured as a policy mechanism for over a century, first introduced after World War One to protect disabled veterans, is that it operates at the level where decisions are actually made i.e. the financial model. Ethical guidelines do not appear in board-level discussions about AI ROI. Regulatory reporting requirements create compliance costs but rarely change strategic direction. A meaningful financial penalty, however, one that shows up on the same spreadsheet as the AI investment itself, changes the arithmetic.
Right now, when a CFO presents an AI automation business case, the model shows: AI investment cost versus labour cost savings. The human suffering that results from the job losses does not appear anywhere on that spreadsheet. It is externalised entirely, borne by the workers, their families, the healthcare system, and the social welfare state. The AI Tax brings that externalised cost back inside the organisation’s financial model, where it belongs.
It will be argued that this makes companies less competitive, that it will slow AI adoption, that it will drive investment offshore. These are the same arguments that were made against disability employment quotas in the 1970s, against minimum wage legislation in the 1990s, and against the introduction of paid parental leave in the 2000s. In each case, the predicted economic catastrophe did not materialise. What did materialise was a gradual shift in how companies modelled their obligations to the humans who loyally work for them.
The question is not whether an AI Tax is perfect. No policy instrument is. The question is whether the current situation in which companies can eliminate thousands of jobs with AI investment, book the productivity gains as profit, and hand the social consequences to the state is acceptable. The historical record on what mass unemployment does to people suggests, very clearly, that it is not.
And finally…that promised Tony Stark lesson that humanity badly needs right now
In the Marvel Universe, Iron Man, Tony Stark is not a hero when we first meet him. He is a weapons manufacturer, brilliant, wealthy, and largely indifferent to the human consequences of what his company produces. He is also, a difficult human being: arrogant, self-absorbed, emotionally avoidant, prone to spectacular failures of judgment. He is not a role model in any way, although if truth be told, he bears a remarkable resemblance to some of the current crop of Tech CEOs.
Anyhow, but then something happens. He witnesses directly, personally, inescapably the human cost of the weapons Stark Industries makes. He watches people suffer and die from his company’s technology. And in that moment, he makes a choice that costs him enormously, that his board resists, that his business partners fight: he stops making weapons. He doesn’t just pivot the marketing. He doesn’t add an ethics checklist to the procurement process. He changes what the company “is” and what it “does”.
He is not a perfect person making this choice. He is a flawed, complicated, sometimes insufferable person making the right choice. And that distinction matters enormously, because the tech industry’s response to AI displacement is increasingly coming down to the same question Tony Stark faced: Now that you can see what this is doing to people, what are you going to do about it?
The executives building AI systems that will decimate millions of jobs are not moustache-twirling villains. Many of them genuinely believe they are building a better world. Many of them are right in part. AI will create enormous value, cure diseases, solve problems that have defeated humanity for generations. The technology is not the enemy.
But 40% of employers globally now anticipate reducing the workforce as AI automates tasks in 2026. When we know what mass unemployment does to suicide rates, to domestic violence, to children’s life outcomes, to democratic stability the question is no longer whether AI can replace workers. It clearly can. The question is whether the people building these systems have the courage to accept responsibility for what that means.
Tony Stark didn’t wait for Congress to pass the Superhero Accountability Act before he changed course. He saw the harm, he accepted personal responsibility, and he used his considerable intelligence and resources to build something that protected people instead of endangering them.
The historical public works programmes that actually worked the WPA, the CCC, the MGNREGA all shared one characteristic: someone in power decided that the dignity and welfare of displaced workers was a problem worth spending serious resources to solve. Not a PR problem. Not a regulatory compliance problem. A human problem.
The legislation being enacted around the world Italy’s prohibition on fully automated employment decisions, the EU’s classification of hiring AI as high-risk, South Korea’s mandated ethics framework, New York City’s bias audit requirements, and now China’s court rulings that AI adoption is a business choice, not a force of nature, and that the cost of that choice cannot be transferred to the worker is the political system doing what it always does eventually: asserting that human beings have rights that markets do not automatically protect.
But legislation is a floor, not a ceiling. It is the minimum that a civilised society demands. The companies building these systems have the resources, the talent, and if they choose to use it the moral imagination to do far more than the minimum.
An AI Tax, modelled on the disability employment levy system that has been proven to work across dozens of countries for over a century, gives boards a simple, unavoidable financial reason to ask a different question. Not “how many jobs can we eliminate with this AI investment?” But “how do we deploy this AI investment in a way that transforms our workforce instead of discarding it?”
That question already has an answer. It just requires someone in the room looking at that spreadsheet to have the same courage Tony Stark showed accepting that the cost of doing nothing is not zero, that it falls on real people and their families, and that it is theirs to own.
Full Disclosure
I founded, own and am the CEO of a company called XTENSOS (www.xtensos.com) that develops various Multilingual AI use-case solutions for enterprise customers. All of our projects are all based on helping our Customers’ existing teams to become more efficient and strategic and never to simply enable staff layoffs. This is a core philosophy of mine, as you can probably tell from this article, and it will never change.
Humans > Machines AI must serve Humans
What’s your take?
Should governments implement my AI Tax idea? Would it change how your organisation thinks about AI investment decisions and workforce design including the related topic of hiring/not hiring college graduates? I’d love to hear your perspective in the comments.
