BAD HOFGASTEIN, Austria — Artificial intelligence poses an “existential” threat to Europe’s health and social security systems through its potential to “completely replace” human work and eliminate the income tax funding base on which they depend, a leading health policy figure has warned.

Clemens Marten Auer, president of the European Health Forum Gastein, used the opening of the summit’s 28th edition in Austria’s Tauren mountains to sound the alarm that AI automation could hollow out the tax revenues that underpin Europe’s postwar welfare systems.

“AI will drastically change, even obliterate, the very foundations of the social contract as we know it today,” Auer said, arguing that the tax base funding Europe’s social provisions requires “total reformation” to preempt the coming transformation of the labour force and collapse of tax revenue.

“AI will transform societies, but may also devalue human labour, deepening inequality and creating instability,” Auer said. “The sustainability of solidarity systems is threatened at its core by this development and cannot be rendered future-proof unless the basis for funding of these systems changes.”

Europe’s social contract—the postwar consensus guaranteeing health, housing and education funded through employment-based taxation—is under attack from all sides as stagnating economic growth and ageing populations collide with budgets strained by war, climate change and digitalisation.

Demographics lead the assault. The EU’s active workforce is projected to shrink by 2 million by 2040, pushing the ratio of workers to pensioners from 3:1 to 2:1 as medical advances extend life expectancy and healthcare costs. Fewer workers mean less tax revenue even as spending on an ageing population rises.

“Society is changing in a positive way. We have lots of technological changes, new drugs, we can treat more diseases,” said Slovenia’s health minister Valentina Prevolnik Rupel. “In developed countries, the number of citizens does not really grow—what grows is the life expectancy. All this is positive, but it brings changes in demand, and needs are growing.”

Those same trends are pushing Europe’s medical workforce to the limit with a projected shortage of 4 million staff by 2030, despite increasing the import of foreign nurses by 67% and doctors by 54% over the last decade, according to WHO figures.

“If we keep the social contributions or the funding unchanged, and on the other hand, we see that the needs are increasing, we have some discrepancies that cause issues in trust,” she said.

‘Learn to speak Russian’

Geopolitical tensions on European borders have intensified competition for scarce funds. Russia’s four-year full-scale invasion of Ukraine has forced a fundamental rethink of EU defence budgets, which stalled in a holding pattern around 2% of GDP in recent decades. Nato allies are now pushing for member states to reach 5% of GDP in defence spending.

The EU has already relaxed budget rules to allow member states to boost defence output, driving a 19% increase in defence spending across the bloc last year.

“If you do not do this, if you would not go to the 5%, including the 3.5% core defence spending, you could still have the National Health Service…the pension system, etc.,” Nato secretary-general Mark Rutte said ahead of a summit in The Hague in June. “But you’d better learn to speak Russian.”

“Sounds a little cynical,” Auer quipped, “but he said it.”

The landmark Draghi report on European competitiveness, released last month, estimates that investment in decarbonisation, defence, economic independence and digitalisation will require an additional 5% of GDP to keep pace with global powers in the US and China, a daunting figure given the bloc’s minimal economic growth and correspondingly stagnant government revenues.

“It’s high time to wake up and to start reflecting these political developments for our sector,” Auer continued. “All these developments will have a massive impact on how we finance, how we organise the level of and the access to health services.”

The 5% calculation in the Draghi report does not include health or pensions, which are largely absent from the document that charts the EU’s growth roadmap and investment priorities. Health and welfare budgets will have to compete for what’s left.

“You need additional money for the investment in infrastructure, that’s the fact. Through social contributions, we just fund the [social] programs,” Prevolnik Rupel said, noting that Slovenia has spent more in the last two years on health than in the previous decade combined to meet rising demand.

“That’s investments in the infrastructure—they are needed because the technological events are so fast,” she explained. “We need new equipment, we need new buildings, we need new spaces, because the protocols of how the patient travels through the systems change.”

Steam engine or hot air? 

The financial pressures on European welfare systems from demographic ageing, geopolitical instability and climate change are clear, as is the looming crisis in health worker shortages. The apocalyptic impact of AI that Auer outlined, however, remains far less certain.

In an essay accompanying his opening remarks, Auer compared AI’s potential to automate work processes to “the invention of the steam engine…and the First Industrial Revolution” in transformative power.

He cited Sam Altman, OpenAI’s chief executive, who predicted ChatGPT “will soon be more powerful than any person who has ever lived,” and Eric Schmidt, Google’s former chief executive, who forecast AI will reach 80 to 90% of top expert skills in mathematics, law, science and programming by 2033. Schmidt maintains one of the largest private AI investment portfolios worth billions of dollars.

Yet nearly all predictions of AI’s looming takeover come from inside the industry itself, where executives and investors have substantial financial stakes in the technology’s success. A significant decline in AI’s perceived value would cost companies like OpenAI hundreds of billions in valuation, and industry figures personally many billions of dollars.

The optimistic narrative has fuelled historic valuations despite record cash burn. The AI industry’s leading companies—Google, Meta, Microsoft, OpenAI, Amazon and Tesla—have spent a combined $560bn on capital expenditure since 2024 and generated just $35bn in return, netting $525bn in combined losses.

Despite the losses, OpenAI reached a $500bn valuation on Thursday, cementing its place as the world’s most valuable start-up.

The lack of profitability has contributed to growing concerns that the AI market might be a bubble, potentially undercutting its perceived revolutionary potential—and threat to solidarity systems in Europe.

AI’s capacity to “completely replace human work,” as Auer wrote, echoing warnings from other AI industry figures such as Anthropic’s Dario Amodei, has scarcely materialised.

A Yale University study published on Thursday found that US labour market metrics show “the broader labour market has not experienced a discernible disruption since ChatGPT’s release 33 months ago, undercutting fears that AI automation is currently eroding the demand for cognitive labour across the economy.”

Meanwhile, AI advancement has stalled as models hit walls in processing power, energy infrastructure, compute capacity and diminishing returns, all while companies have yet to identify profitable use cases for the technology.

A Massachusetts Institute of Technology report in August found that 95% of companies experimenting with AI are not making money from it, despite widespread promises from corporate leaders about efficiency gains.

Large language model technology is currently built on burning cash and the environment. Data centres consume more energy annually than all but 16 of the world’s 192 nations, with consumption set to double by 2030.

Global AI demand for water to cool servers and microchips is projected to require a supply equal to the total annual water withdrawal of countries like Denmark by 2027.

Continuing AI scaling and power-hungry data centres are eating into gains from green energy sources such as solar and wind, forcing gas, oil and coal plants back online to meet demand as air, noise and environmental pollution from digital waste threaten the health of nearby communities.

AI leaders including Altman have acknowledged that an energy breakthrough, likely from nuclear fusion, will be required if AI is to continue scaling. Google and Microsoft have started investing in nuclear technology, but a scalable breakthrough in the field is still a long way off.

“Naturally, historic predictions are difficult, especially when it comes to developments and employment in the labour market,” Auer acknowledged. “I do not want to oversimplify the issue here.”

AI’s healthcare promise 

If AI does fulfil its transformative potential, healthcare is seen as one of the most promising frontiers. Yet here too, the gap between promise and reality remains wide.

Nearly two-thirds of physicians surveyed in the United States reported using AI in 2024, a 78% jump from 38% the previous year, according to a study by the American Medical Association. The usage, however, was primarily for administrative tasks such as documentation of billing codes, medical charts, visit notes, discharge instructions and translation services rather than medical or scientific breakthroughs.

Slovenia’s experience illustrates the challenge facing health systems. The government issued an open call for AI healthcare proposals and received 52 potential solutions for a country of just 2 million people.

“So it’s like a flood of artificial intelligence applications, and we have to decide which ones to take on—and it’s very difficult because we don’t know,” Prevolnik Rupel said. “We don’t have enough evidence yet on which solutions work and which don’t, and there are many.”

The government decided to pilot the most promising submission, an artificial intelligence program for scheduling healthcare workers’ shifts. Like the administrative applications in the AMA survey, the use case was organizational rather than medical, as doctors remain wary of trusting AI with decisions that could impact patients.

“Because it was an artificial intelligence logical program, it was more fair than when it was a nurse who would say, you do this,” she said. “Everybody was happy, and the level of satisfaction increased, and actually, the absenteeism among the healthcare workers fell in that hospital.”

The proliferation of untested AI applications has raised questions about how to regulate the technology safely. But experts say existing regulatory frameworks can be adapted rather than reinvented.

“When we started regulating medicines, I think 70 years ago, more or less in the European Union, was one of the first success stories,” said Natasha Azopardi-Muscat, division lead for WHO’s Europe office.

“We said medicines have to meet certain criteria: quality, safety, efficacy. And then we went further. We said that they have to be better than the alternative,” Azopardi-Muscat explained. “We can use the same [standards] for AI… We need to have a very clear system of evaluation.”

Despite the gap between AI predictions and current reality, Auer argued that political leaders are ceding crucial decisions about society’s future to private investors.

“If the automation stimulated by AI leads to a broad reduction of costs, meaning labour costs, meaning a significant loss of jobs,” he said, decisions about “how much capital to invest in development are made by financial investors, and not in parliaments or cabinet tables of governance.”

“The political call of the economic elites to strengthen Europe’s innovative power as forcefully as possible will come to nothing if at the same time we ignore the political concerns that innovations of the tech industry have the potential to dehumanize economic activity and weaken human labour,” he said.

“We must discuss and politically initiate steps to protect the solidarity systems in the long term and ensure the social and economic participation of as many segments of the population as possible.”

Image Credits: CC.