Big Tech should not get to decide unilaterally which AI risks the public will bear

[PARIS] The US midterm elections were a clarion call for change.

Affordability concerns, the war in Iran, and corruption all played a part in the Democrats’ sweeping victory, signalling widespread dissatisfaction with America’s direction, anger over unfair economic outcomes, and anxiety about a future that seems destined to be shaped by artificial intelligence.

The technology is being developed and deployed in a manner that has worsened economic insecurity and aggravated the public’s sense of a loss of control. The most visible element concerns data centres and the pressure they put on electricity bills and local water supplies, but the issue is much broader.

Voters increasingly worry about growing inequalities of income and opportunities, and they are right to do so.

Over the past few decades, the gains from growth have flowed disproportionately to those at the top. US Federal Reserve data for the second quarter of 2026 show that the wealthiest 1 per cent of US households held 32.5 per cent of the country’s net worth, while the bottom 50 per cent held just 2.3 per cent.

The tech sector exemplifies this concentration. Its multi-trillion-dollar companies and first trillionaire (Elon Musk) are not contributing anything close to their fair share of taxes.

According to the Institute on Taxation and Economic Policy, Amazon, Alphabet, Meta and Tesla (which it categorises as the “four big pro-Trump tech companies”) paid federal corporate income taxes equal to just 4.9 per cent of their US profits in 2025. At the 21 per cent average rate, their tax bill would have been US$51 billion higher.

Meanwhile, public perceptions of undue industry influence in government have grown – again, rightly so.

The Trump administration’s unwillingness to regulate the AI sector, even after clear evidence on many grounds and serious hacking incidents, and Musk’s decision to pour almost a half-billion dollars into Republican candidates’ campaigns in 2026, merely added fuel to the fire.

Then there are voters’ concerns about AI and jobs, discriminatory algorithms and addictive digital platforms that harm young people. These issues still have not received the regulatory and policy attention they deserve; nor have the risks posed by hyper-paced AI development, including the possible loss of control over the most powerful systems.

Many Democratic candidates astutely addressed these concerns head-on, promising to restore people’s sense of agency in the face of a technological revolution that is barrelling ahead. Texas Governor-elect Gina Hinojosa called for a moratorium on data-centre construction until legislative protections are established.

US Senator-elect Abdul El-Sayed called for a Food and Drug Administration-style agency to evaluate frontier models before they are released, and voiced support for a 50 per cent public equity stake in the largest AI companies.

And many other candidates pointed to the administration’s support for data centres as proof that the Republicans prioritise corporate interests over those of workers and local communities.

Now that the Democrats will hold sizeable majorities in both chambers of the US Congress, they have an opening to push for long-overdue legislation to steer AI’s development towards the common good.

The Big Tech companies have far too much power to decide unilaterally which risks the public will bear. That needs to change. In no other economic sector are private companies permitted to assess their products’ risks and decide for themselves whether they are safe to release.

One important immediate step would be to pause development of the most advanced AI models until lawmakers have established a clear liability regime to hold companies accountable when things go wrong.

The recent incidents in which AI agents hacked other companies, and even government ministries, had no consequences (though Australia has at least launched an investigation). A human employee who had done the same thing would face criminal charges.

To ensure that any new rule or regulation works as intended, we also need well-resourced independent institutions for oversight and enforcement. Assessing the risks society faces is the proper responsibility of elected officials.

Fortunately, the new Congress will not be starting from scratch. After the recent hacking incidents, several members of Congress (from both parties) introduced bills to address some of the downsides of AI, and all converged on the core issues of safety, accountability and greater public control, though they differ on the means to uphold these principles.

Senator Bernie Sanders, for example, has called for a Cabinet-level office that would oversee AI development. Senators Mark Warner, Brian Schatz and Andy Kim want to create a Department of Commerce-based safety board with access to models before release.

And Senators Dick Durbin (a Democrat) and Josh Hawley (a Republican) would classify AI systems as a “product”, thereby facilitating traditional product-liability claims. Senator Mark Kelly proposed to create an AI Horizon Fund to support labour transitions, financed through taxing big companies on excess profits, digital advertisements and AI usage. Sanders also proposed the creation of a wealth fund with similar purposes.

Of course, President Donald Trump may resist these proposals, and he has already threatened to veto any AI bill he disagrees with. But if he reads the writing on the wall, he will see that putting the technology on the right track is good for his legacy.

In fact, Congress could build on Trump’s June executive order, which included provisions for voluntary government access to frontier models. These provisions, along with the safety commitments that industry leaders made in September, should become legally enforceable obligations.

At the same time, the Trump administration can use its AI dialogue with China to pursue cooperation on the biggest risks.

America’s voters have given the country a fresh start. Their representatives now have an opportunity to act. If they waste it, another one may not come along in time to avert a catastrophe. PROJECT SYNDICATE

The writer, co-chair of the Task Force on Inequality and Social-Related Financial Disclosures, is a former assistant director-general for social and human sciences at Unesco, where she oversaw the development of the Recommendation on the Ethics of AI; and a former OECD chief of staff and sherpa to the G20, G7 and Apec