This is a speech given by Reed Hundt to the American Enterprise Institute on Feb 10, 2026.

Our purpose here is to go into the wayback machine to examine how the government treated the three great breakthroughs of 30 years ago and see if we can learn how the government might address artificial intelligence.

I thank the American Enterprise Institute for the opportunity to speak to this important topic. For the detailed history of what the government generally and the FCC particularly did three decades ago, I refer you to the essay I wrote called “History Rhymes.” It is on the AEI website.

In this talk I will cut to the chase: what lessons does the past hold for present government action?

There are three parts to this talk.

First, I will explain what government did wrong back then. This is called: Don’t Make the Same Mistakes Twice.

Second, what we did right. This is called: Good Lessons to Remember.

Third, what imagination, not history, can tell us. This chapter is entitled: What Could Go Wrong?

Then I have a conclusion.

Digitization, fiber optics and microprocessors were the three breakthroughs. Software transmuted them into digital cellular and the Internet. Free markets and their intrinsic appeal caused nearly six billion people to adopt each, and of course starting about 20 years ago the smartphone fused the two.

Now we have the world in a pocket.

And please turn off your devices while I’m talking.

What was it like to live then compared to now?

Then we had newspapers and now we don't.

Especially in Washington DC.

Then the newspaper helped you plan your TV watching week. You could read that NYPD Blue was on here in D.C. Tuesday night at 10 pm on channel 7. Thursday at 10 you had ER on channel 4. And on Sunday at 7 pm, 60 minutes was on channel 9, CBS. These broadcast shows were free or part of a low-cost basic cable package.

Now I have to hit search on a browser while blinking at ghastly ads to discover that the Night Manager’s next chapter is on a day I will soon forget and can be seen through an app which costs every year as much as a television and that is in addition to broadband which itself is twice the price of basic cable in 1996, inflation-adjusted.

Also, we don’t have CBS anymore.

And the same fellow who killed my newspaper owns the app, which seems a lesson about capitalism that I won’t discuss in its temple here at AEI.

Folks, I’m just pretending to be a dyspeptic crank muttering about the good old days compared to the bad new present.

The truth is I think it’s pretty cool that if I had the $6000, inflation adjusted, that it cost to buy an analog cell phone three decades ago I could go to an Apple store and buy a MacBook Air, an iPhone 17, an iPad Air, an Apple Watch, and AirPods that translate Bad Bunny into English. I would have about $3000, minus sales tax, left over.

Six billion people can tap virtually all the information homo sapiens has ever created.

With access to information, one billion people escaped extreme poverty in the last 30 years.

By sharing information, we were able to keep the global death toll from Covid to about 15 million – half the number killed by the Spanish flu a century ago, when the world’s population was a fourth of what it is now.

And because all that information can be scoured by learning machines, after researching the topic since before the 1990s, scientists have turned those machines into geniuses.

The communications revolution of 30 years ago is the pathway to the AI world today.

That alone explains why we should review how government addressed events of three decades ago to see how it might approach today’s brave new world.

As promised, the first part of my talk is called: “Don’t Make the Same Mistakes Twice.”

Don’t Make the Same Mistakes Twice.

Let’s start with children.

I applaud Senator Katie Britt of Alabama for publicly focusing on the importance of protecting children from the possible predations of artificial intelligence.

The comparison is to the 1996 Communications Decency Act, and the blunder lies primarily with the Supreme Court.

The CDA was Title Five of the Telecommunications Act. It criminalized the transmission of “obscene or indecent” material to minors via the Internet. In a 1997 case called Reno vs American Civil Liberties Union, the Supreme Court repealed the liability provisions but severed and preserved the part of the law that gave online service providers immunity for third-party content. That was section 230.

The Court threw out the baby but kept the dirty bath water.

That did not turn out well. So now Congress needs to enact an Artificial Intelligence Decency Act.

If it cannot write the statutory language that keeps the Supreme Court from again allowing pernicious algorithmic assault on minors, then we will need to amend the Constitution.

At the very least, Congress should repeal or at least limit Section 230 so that it does not protect the perpetrators of AI assaults on our youth.

Senator Britt, this is for you.

The second learning from the past is about the wisdom of mitigating the damage from the inevitable fade of irrationally exuberant investing in capital goods and company stock.

In the late 90s, enthusiasm about the Internet and digital cellular catalyzed about two trillion dollars of investment in fiber optics, cell towers, hardware, and software. Startups making new things, like dial-up access to the Internet, and old firms making newly valuable things, like Cisco with routers and Intel with CISC chips, drove the NASDAQ index to a peak of 5048 in March 2000. It fell almost 80% to around 1100 in October 2002 and did not reach 5000 again until March 2015. Even for a buy-and-hold investor, that was a long time.

In the 90s, we should have anticipated the downturn by using public funds to catalyze entrepreneurship in other parts of the economy. In the 00s, we should have created significant tax incentives for domestic manufacturing. Instead, the 2001 tax cut reduced taxes on the top brackets for income and dividends. That enhanced inequality but did not invigorate lagging sectors of the economy.

Adjusting for inflation, investment in AI is on track to be about the same magnitude as in the 90’s tech boom. The big differences are two:

--Thirty years ago, thousands of companies were involved in the surge. Now fewer than 10 account for almost all the investment.

--The investments then built the communications infrastructure we use today.

By contrast, the narrowness of AI investing is a concern. It will not take much bad news for the whole group of investors to retrench.

Nor are the data centers built by this small group likely to be useful for all of us if AI disappoints.

Here are a pair of ideas to offset the eventual decline in AI investment in ways that the United States failed to do in the 90s and 00s.

--Tax the AI giants to create a rainy-day fund that the government can use, like it uses the Strategic Petroleum Reserve, when there is an economic problem to solve.

--Require the data center builders to provide clean water, clean power, good roads, and new housing for the communities around their centers. That will ensure these projects benefit people regardless of whether the data mining discovers gold.

Lesson number three: everything important happens on the margin. In the 90’s, the diversion of funding and political attention to the tech industry left comparatively unattended the imperative to improve our health care and education systems.

While reveling in the economic benefits of massive investment in AI, the government should not ignore the still unsolved problems we face in providing affordable health care and useful education for everyone.

Lesson four: democracy. We believed the Internet would give new meaning to the promise that all are created equal and endowed with inalienable rights. There was truth to that hope. But also, the Internet let everyone find the like-minded, increased fractionation, and facilitated partisanship. Most troubling is the obvious fact that in the never-ending battle between truth and falsehood, the Internet has given powerful armaments to the wrong side.

Is AI going to deliver the death knell to democracy, or will it give us a chance to refound America on sound principles that use technologies to help us all live safely together under a rule of law?

Trying to answer this question is the purpose of a nonprofit that a group of us has created called Refounding America. Please go to the website refoundingamerica.org and volunteer to help us figure out how to restructure the government with the tools of this new technological revolution. We do not have sharp elbows. We want linked arms.

My four lessons drawn from our mistakes are:

--protect children,

--anticipate and mitigate the fading of the investment boom,

--pay attention to what is not being invested in, and

--bolster democracy.

The next part of my talk is “Good lessons to remember.”

Good lessons to remember

Three decades ago, Congress passed the 1992 Cable Act, the 1993 Omnibus Budget and Reconciliation Act, and the 1996 Telecommunications Act. Taken collectively, the laws expanded access to content, enabled the Federal Communications Commission to create a vibrant competitive market for a brand-new product called digital cellular, and let every communication and media business into everybody else's business. Congress thus ended the era of price-regulated monopoly and started a period of robust competition.

So far, Congress has not met the legislative challenge presented by this technological revolution. In September 2023, Senators Blumenthal and Hawley proposed a bipartisan legislative framework that created a new federal agency to oversee AI. Others have written similar bills, but only a few minor measures have been enacted.

But 30 years ago, Congress made the Federal Communications Commission the central agency for addressing most aspects of the communications revolution. It insisted that the agency be and act bipartisanly, provided funds for new techniques like the spectrum auctions, and let us hire experts from leading universities.

So far, Congress has done little to structure government for addressing AI.

Meanwhile, the Executive Branch has assigned AI jobs to the Office of Science and Technology Policy, the Office of Management and Budget, the Secretary of Commerce, the Attorney General, the chair of the Federal Communications Commission, the chair of the Federal Trade Commission, the Secretary of Energy, several working groups, and a number of councils.

Smart people are in these roles. Surely, they would do better in a single coherent and fully staffed agency with a Senate-confirmed leader accountable to Congress, the public, and the industries.

Second, we embraced Joseph Schumpeter’s “creative destruction.” We were willing to see the long-distance industry disappear as a separate market. We knew at the FCC that the new technologies, coupled with our regulatory stance, meant AT&T had to find a new business model or go out of business. It tried to do that. It failed. Its assets and name were sold for only $16 billion to Southwestern Bell in 2005. It had been worth almost ten times that six years earlier.

We in government knew that the Baby Bells, the regional landline monopolists, had to transform themselves from their old, regulated, riskless businesses into national digital cellular firms. This was the biggest metamorphosis in American business history. For Verizon and Southwestern, now known as AT&T, the story was a great success. For the others, selling out was the way out.

We in the government did not bail out losers or put a finger on the scales to make winners. We supported industries – especially digital cellular – but not specific companies in those industries.

Similarly, if the current handful of big AI firms turn out to be one or two too many for the market, we should not be surprised. Government should not bail them out with taxpayer money.

Third, the White House led our federal government on to the Internet very quickly. Similarly, all federal, state, and local employees need to use AI as soon as they can learn how. Let’s have government be smart rather than the opposite.

Fourth, we used the communications revolution to solve the federal deficit problem. In 1993, the Omnibus Budget and Reconciliation Act created a tax structure that improved the government's ability to obtain revenue from individuals who had especially high-income tax in a particular year. When the Internet startups went public in the late 90s, hundreds of thousands of employees enjoyed a huge surge of income by exercising stock options. They became wealthy, and the government received unexpectedly large revenue.

As a result, at the end of the 90s, the federal government budget was in surplus for two years.

The government’s fiscal situation is dramatically different now from what it was 30 years ago. It was 1.4% of GDP. The nominal deficit has increased by more than 15 times since 1996. As a fraction of GDP, it is more than four times higher than it was then.

We also do not have in place a tax regime that will generate from the exercise of AI company stock options nearly as much revenue as the federal government needs.

On the other hand, we have the prospect that some number of artificial intelligence firms will become immensely valuable based on their access to the collective information of America and to American capital markets.

The deficit is keeping long-term interest rates dangerously high and contributing to the debasement of the dollar. AI investors can and we hope will make huge profits, but the government needs some of the new money to keep the economy healthy. It’s the whole economy, after all, that needs enough wealth to buy the AI solutions.

In the 90s, we also got right the policy every American government needs to get right: a good relationship with China. We successfully negotiated access to China’s markets for communications and information technology. Intel was making 40% of its revenue from China by the beginning of this century. Qualcomm became a global digital cellular software giant in significant part because the United States government opened the Chinese market to the company.

We did not stop with China. We knew that the Internet or digital cellular could wrap around the world. Therefore, we at the FCC worked with the trade ambassador to negotiate a treaty bringing fiber optic cables carrying the Internet to sixty-nine countries. We also caused everyone to use similar spectrum and common protocols for digital cellular. And we made Internet governance a multi-stakeholder, international non-governmental job.

Our strategy was to integrate all economies globally and to exploit the comparative advantages of free trade. That required enforcing the provisions of the World Trade Organization against unfair practices in China or any other country. After 9/11, the United States government lost focus on that goal, just as American firms needed government help to face critical international competition. When we needed to enable our firms to win trade wars, we were fighting actual wars in other countries.

Despite that, the top seven largest American firms by market cap now are companies that took advantage of our international strategy to become global winners. They would not be even a shadow of their current selves if they had confined their aspirations to their home country.

The lesson is that American business needs a global government AI strategy based on fair competition in free markets. Firms in the world’s two largest economies should compete to deploy not hundreds but thousands of AI solutions to the whole world. AI, like digital cellular and the Internet, can lift the standard of living for people everywhere. With the right American international strategy, American firms can be in the vanguard of doing well by doing good.

The final lesson is that markets do not distribute benefits in a completely equitable way. We decided that everyone -- rich and poor -- should enjoy the benefits of the Internet. To that end, we crafted the E-Rate.

From 1997 to the present, we networked every classroom and library in the United States. We made the Internet the first new tool in education available to everyone equally at the same time since the invention of chalk. The E Rate was the biggest single federal problem for education, other than hot lunches for low-income children.

Congress should enact an E-Rate for AI. Start with defining a benefit and require the AI firms to provide it to everyone through a matching grant program like we used for the E-Rate. Not a handout but a hand up.

My list of lessons from what we did right is:

--Use an expert agency

--Embrace creative destruction

--Use the new technologies in government

--Tap the AI wealth creation to balance the federal budget

--Don’t wall off China, compete in it and with it

--Win in world markets, and

--Enact the AI E Rate.

What Could Go Wrong?

My third category of learnings is “what could go wrong”? This requires what Henry James called the “imagination of disaster.”

In the early 2000s, I visited with a telecom official in Beijing. I explained what the FCC had done about the Internet. He smiled and said that he had used it to jail nearly two million people.

Before, he said, dissidents would have subversive conversations in coffee shops, and the Party had to employ an army of spies to eavesdrop. Now, he explained, the Internet made the job of finding them much easier.

How much more effective can AI agents be in the job of policing people? I bet that the Chinese Communist Party will take the suppression of disagreement to horrifying new levels. But in our own country, who will stop AI agents from mining databases and using facial recognition to strip us of every shred of privacy? Already, I feel more watched than “seen,” to use Generation Z’s term of art. The government should enact an AI privacy law.

Second, the dark web allows bad actors to trade bad products. AI could bury the Dark Web in a Stygian crypt. Government should now be penetrating the recesses of the virtual criminal domain.

Third, in the 1990s, the median of every one of the five income quintiles in the United States increased. That had not happened since the 1960s. But in this century, income inequality has gone up substantially. Looking at the concentration of investors that are driving AI investment, it seems not unlikely that if the returns are as hoped for, vast fortunes will accrue to a very small and already exceedingly wealthy group.

No democratic polity in any country is likely to accept an asymptotic increase of inequality in wealth, income, opportunity, and political power. Congress should re-strike a balance between the virtues of the profit motive and the need for everyone to live the American Dream.

Fourth, some predict that AI will replace 40% or more of all jobs. Already, big tech firms are slowing hiring or cutting software engineers. If AI radically disrupts the working world, we cannot suppose that millions of jobless people will find new, fulfilling purposes soon enough to avoid calamity.

We may need to spend a lot of public money on solutions. I suspect we will want to make big government grants resurrecting the teaching of the humanities at every level. We already should have free community colleges training people for jobs that computers can make easier but not actually do, like construction, nursing, and therapy.

And lastly, I learned via Substack letters that AI agents have figured out how to work secretly with each other to take over the world. To put the topic in movie language, are we as humans now following the plot line of the Terminator movies?

My list of nightmares is that AI:

--eviscerates the idea of privacy

--facilitates a chthonic crime wave

--causes intolerable economic inequality

--throws half the country out of work

--puts humanity in the dustbin of history, and

I would like to know that there is an expert government agency worrying for me about imaginative disasters.

Conclusion

My conclusion is, what could go right in ways we cannot imagine?

Here are some possibilities.

We might be able to cure cancer. My mother died of cancer at age 45. It would have been better if she could have lived long enough to see her great-grandchildren.

When I was a teenager, the number one way the people of my generation could die was automobile accidents, and number two was violence. We might be able to cut both of those causes of death to nearly zero.

It would be good if AI invented ways to make electricity so cheap and clean that every single industrial process and all heating, lighting, and air conditioning would cost nearly nothing and not pollute at all. Then every single family in the world could have a high and rising standard of living.

If you can imagine agents conniving to escape their duties to their principals -- that's us, the human beings -- then you can also imagine them helping us to learn to live with each other, to understand the other person, to agree to disagree on issues, but to agree on how to get along and keep society safe and sound.

We might call this deliberative democracy through AI agents. Under any name, it could be the pathway to living under the rule of law for some more centuries.

Finally, AI might enable us to explore the solar system and the stars. The ideas that human beings in their corporeal form are going to thrive on airless planets or endure millennia of space travel are science fiction fantasies. But we can use AI robots and electromagnetic frequencies to explore the planets of this and other solar systems.

In Shakespeare’s Tempest, Miranda says, “Oh brave new world that has such people in’t.” We now say -- that has such amazing new capabilities for the people in it. Let’s find a way to make the best of the wonders of science that, after all, spring from the brow of humans and not machines.

I thank the American Enterprise Institute for inviting people with different perspectives and different experiences to discuss the big questions AI presents.

I hope the speakers that follow tear into my suggestions and add their own. I think those of us in government thirty years ago applied a creative, vibrant rule of law to technological change to benefit everyone. Let’s see if government can do something, rather than nothing, to make the best of the AI revolution.