AI Has A Public Trust Problem

The people building artificial intelligence keep telling us how profoundly it will disrupt our lives. Americans are listening — and increasingly they don’t like what they’re hearing.

Political Sarcasm, Created Using AI

Our morning content is free for all subscribers and guests! Usually our afternoon content is behind a paywall… If you are not yet a paid subscriber, consider upgrading — you are missing a significant portion of what we produce each week, and your support makes it possible. If you want to get it all and support my efforts, please consider a paid subscription! There is significant exclusive content below the paywall, after this morning’s column.

Most of our content is available on our podcast feed. So, Does It Matter? SPOKEN. Which is available on your favorite podcast app. You can listen to it right here as well.

🕒 7-minute read

---

Listen To The People Building It - In Their Own Words

Is the rapid development of artificial intelligence going to be extraordinarily disruptive to the way Americans work and live? Don’t ask me. Ask the people building it.

Elon Musk, CEO of Tesla and xAI, has predicted a future dominated by artificial intelligence and robotics in which “probably none of us will have a job.”⁠ Anthropic CEO Dario Amodei has warned that AI could eliminate half of all entry-level white-collar jobs and drive unemployment to between 10% and 20% within one to five years⁠. “Most of them are unaware that this is about to happen,” Amodei said of policymakers. “It sounds crazy, and people just don’t believe it.”

Amodei has an even more striking way of describing the strange future he thinks may be coming: “Cancer is cured, the economy grows at 10% a year, the budget is balanced — and 20% of people don’t have jobs.”⁠

Microsoft AI CEO Mustafa Suleyman thinks enormous changes could arrive even sooner. “I think we’re going to have a human-level performance on most, if not all, professional tasks,”⁠ he said. After specifically mentioning lawyers, accountants, project managers and marketing professionals, Suleyman predicted that most of those tasks could be fully automated by AI within the next 12 to 18 months.

OpenAI CEO Sam Altman has been similarly blunt about customer-service jobs. “Some areas, again, I think just like totally, totally gone,”⁠ he said, describing AI agents capable of handling entire customer-service interactions without transfers or phone trees.

These aren’t warnings from anti-AI activists. They are predictions from some of the most important people building the technology. Maybe they’re right. Maybe they’re exaggerating. But Americans are listening, and increasingly they are worried.

According to Pew Research Center⁠, 51% of Americans say they are more concerned than excited about increased AI use, while only 11% are more excited than concerned. On employment, 56% are extremely or very concerned that AI will eliminate jobs, compared with only 25% of AI experts.

Disruption Sounds Different From The Outside

Major technological advances have always been disruptive. Mechanization transformed agriculture, automobiles devastated horse-related industries, computers eliminated clerical jobs and the Internet destroyed entire business models while creating new ones.

On balance, innovation has made us enormously more prosperous. I am an optimist about technology, and I don’t want politicians freezing it in place to protect existing jobs. But understanding disruption in the abstract is very different from being the person whose livelihood may be disrupted.

If you’re a venture capitalist hearing that AI may automate most professional tasks, you may hear an extraordinary investment opportunity. If you’re a 48-year-old accountant with a mortgage and kids approaching college, you may hear something else entirely. In Silicon Valley, “disruption” is practically a compliment. The rest of America doesn’t necessarily experience it that way.

Now AI Is Coming To Your Neighborhood

Artificial intelligence requires a staggering amount of physical infrastructure. Data centers need land, electricity, water, transmission capacity and, in many cases, new power generation. But the industry’s political agenda goes well beyond infrastructure. The federal government has embraced accelerated permitting and financial support including loans, grants and tax incentives⁠ to help build AI infrastructure, while its national AI framework recommends federal preemption of many state AI laws and limits on state efforts to hold developers responsible for third-party misuse of their models⁠.

There may be reasonable arguments for some or all of those policies. Some aspects of AI regulation clearly lend themselves to national standards. But Washington should not preempt the field entirely. Federalism allows states to experiment and learn from one another rather than locking everyone into a single approach.

Whatever the merits of each policy, look at the cumulative ask: government assistance to build quickly, protection from significant state regulation and limits on some legal liability — while the industry’s own leaders are telling Americans that AI may eliminate huge numbers of jobs and profoundly disrupt everyday life.

AI is no longer something happening invisibly inside your laptop. It is showing up in communities, legislatures, courts and electric bills.

A Gallup survey⁠ released this year found that 71% of Americans would oppose an AI data center in their area, with 48% strongly opposed. For comparison, opposition to a nearby nuclear power plant was 53%. Americans are significantly more resistant to an AI data center than a nuclear plant.

That resistance is already having consequences. Data Center Watch⁠ says at least 75 projects representing roughly $130 billion in investment were blocked or delayed during the first three months of this year. More than 300 state bills dealing with data centers were introduced during the first six weeks of 2026 alone.

Much of the opposition is about exactly what people say it is about: electricity and water consumption, pollution, quality of life, utility bills and taxpayer subsidies. Those are real issues. But I wonder if something else is going on as well.

I haven’t seen polling that tells us how much opposition to data centers is really a manifestation of broader apprehension about artificial intelligence itself, but I wouldn’t be surprised if it is significant. When Americans are repeatedly told that AI could eliminate huge numbers of jobs and profoundly disrupt their lives, the massive data center proposed down the road becomes more than another industrial development. It becomes the physical embodiment of the darker side of the AI revolution they keep hearing about.

Maybe future polling will prove me wrong. But it seems implausible that Americans can be increasingly apprehensive about AI in one poll and overwhelmingly opposed to the infrastructure needed to power it in another, with no relationship between the two.

AI Builds Political Muscle

The AI industry understands that public policy will shape its future. Companies want to build data centers faster, secure enormous amounts of electricity, limit regulations they believe inhibit growth and preserve freedom to develop their technology. They also argue that these policies help maintain America’s technological advantage over China — a national interest that happens to align quite nicely with their own.

And they are putting serious money behind that effort. The pro-AI political network Leading the Future has raised more than $125 million⁠ for the 2026 election cycle to support candidates favorable to AI development and oppose those it considers hostile. There is nothing wrong with that. Every major industry plays politics when government decisions can dramatically affect its future.

But there is a disconnect. The industry is rapidly becoming sophisticated about influencing politicians while proving much less successful at persuading the people who elect them. Spending millions supporting candidates friendly to AI doesn’t solve the underlying problem if 71% of Americans don’t want an AI data center built near them.

Political influence can win important battles. Public opposition can eventually determine the war.

What About The Rest Of Us?

This is where the industry’s challenge becomes more fundamental than public relations. AI companies don’t simply need to find better ways to sell Americans on what they are doing. They need to earn the confidence they are asking the public to place in them.

If Dario Amodei genuinely believes his technology could cause 20% unemployment, I want him to say so. The answer isn’t for AI executives to conceal what they think is coming because the public might react badly. But candor has consequences, particularly when the same industry predicting enormous disruption is asking for substantial public support and political latitude to continue developing at breathtaking speed.

So what exactly is the bargain being offered to the American people? If AI will be as transformative as its creators say, how much disruption should we expect? What is the industry doing about the upheaval its own leaders predict? Who bears the enormous infrastructure costs, and what responsibility will these companies accept when things go wrong?

Most importantly, what does the average American get out of this transformation? The industry has become extraordinarily good at explaining what artificial intelligence will be capable of doing. It has been considerably less successful at explaining why ordinary Americans should want the future it is building.

So, Does It Matter?

Apparently, at least one of the most important people in artificial intelligence understands the problem. Nvidia CEO Jensen Huang has criticized the rhetoric coming from his own industry, saying, “I don’t know why AI companies are trying to scare us.”⁠ More recently, he put it even more sharply: “Warning people is one thing. Scaring people is a completely different thing.”⁠

The same predictions that demonstrate AI’s extraordinary value to investors can sound terrifying to people whose lives may be disrupted by it. AI may ultimately increase productivity, cure diseases, create new industries and make Americans considerably wealthier. History gives us plenty of reason for optimism.

But history also tells us something about politics. If Americans increasingly conclude that AI is something being done to them rather than something being built for them, eventually the political system will respond. Data centers will become harder to build. Politicians will discover that attacking AI gets votes. Government will face pressure to intervene. And the industry could end up inviting precisely the heavy-handed government involvement it would rather avoid.

The people building AI cannot expect Americans to support the freedom and infrastructure necessary to transform their lives without first convincing them that the transformation is in their interest.

Artificial intelligence doesn’t just need technology, capital, electricity, or political influence. It needs public legitimacy.

---

More Content Below for Paid Subscribers

When AI Becomes a Guide to Mass Harm

A new Wall Street Journal investigation examines what happens when increasingly powerful AI chatbots are asked to help make biological weapons, poisons, or plans for mass-casualty attacks.

The story raises the same question of public trust discussed above. Among the Journal’s findings:

• Some chatbots provided credible information about biological weapons and poisons.

• Experts concluded that certain responses could pose a genuine danger.

• Suspicious users may be banned without being reported to law enforcement.

• Determined users can sometimes circumvent AI safety restrictions.

• Tighter safeguards can interfere with legitimate scientific research.

• Congress is considering reporting requirements and greater federal oversight.

Paid subscribers can continue below for a preview and a gift link to the complete Wall Street Journal investigation. (I pay $39/month for a WSJ Subscription so you don’t have to…)

The Journal Found Something Disturbing…

In a new investigation, Wall Street Journal reporters Georgia Wells and Amrith Ramkumar examine the biological capabilities of leading AI models—and what can happen when people try to misuse them.

They report that hundreds of users asked an upgraded version of ChatGPT how to make or deploy biological weapons and poisons. In some cases, it supplied detailed instructions. Experts who reviewed certain conversations concluded that some answers were accurate enough to be dangerous, according to the Journal.

OpenAI banned the accounts involved but didn’t alert law-enforcement authorities, the Journal reports.

That exposes a gap in federal law. AI companies generally aren’t required to report users who ask for help making weapons or planning attacks. The companies largely decide when a conversation warrants contacting authorities.

OpenAI says it trains its models to reject harmful requests, tests them before release and monitors attempts to obtain dangerous biological information. It also says it contacts law enforcement when it identifies an imminent and credible threat. Other leading AI developers have safeguards of their own.

Those protections aren’t foolproof. Researchers told the Journal that persistent users can sometimes circumvent them during extended conversations. Open models present another challenge because users may modify them and remove their restrictions.

Blocking every question involving dangerous biological material isn’t a workable solution. Public-health officials, medical researchers and drug developers may need similar information. A model that rejects every sensitive question might be safer, but less useful during an outbreak or scientific investigation.

That is the industry’s dilemma. Companies want models capable of helping researchers solve difficult biological problems. Those same capabilities could help people with very different intentions.

Members of Congress have proposed legislation requiring AI companies to report certain dangerous threats. Another proposal would allow the federal government to shut down models deemed too dangerous.

This is no longer merely a debate about what AI might someday do. Companies are making these decisions now, mostly under rules they established themselves.

As you can tell, this kind of stuff feeds right back into the point of the main column above — AI companies have a real challenge on their hands.

Paid subscribers can read the complete investigation by Georgia Wells and Amrith Ramkumar through this GIFT LINK.