For years, I have looked for any
excuse to redirect a conversation toward the impacts of artificial
intelligence. But when I launched my campaign for Congress last December, I
committed to not bringing up the topic for the first few months. I wanted to focus
on the skyrocketing costs of groceries, health care, rent, and childcare to
build the trust needed to raise concerns about AI.
I was wrong.
In my first week on the trail,
voters asked me about AI at every event we held. With intensity in their
voices, parents were asking me how to protect their kids from addiction and
self-harm and to protect their aging parents from increasingly sophisticated
scams. Educators are wondering what skills are still worth teaching kids when
“Nafta on steroids” is hurtling toward us. More than a few pastors
have suggested the devil is at work through AI.
And, yes, voters are raising more
and more concerns about the potential for existential risk (“x-risk”). At a
farmers market this summer, a 73-year-old watermelon grower from Halifax
County, during a conversation about the suffocating costs of fertilizer and
diesel, said, “Yeah, this war is killing us. But what about these rogue
bots? Seems like they might just kill us all.” At a town hall in
Pittsylvania County this month, one-third of all questions were about AI—over
half, if you include questions about data centers.
The American people are hungry for
a serious national strategy on AI. So far, neither party has offered more than
table scraps.
The most obvious explanation for
this gap is to follow the money. Big Tech has built a nine-figure war chest
full of carrots and sticks. When state Assemblyman Alex Bores wrote New York’s
AI safety law and ran for Congress, tech giants promised to make an example of
him (their words!), and Leading the Future and affiliated PACs spent more than
$8 million to beat him in his primary. They have already shown that they will
bankroll campaigns of those who bend the knee. Certain tech titans spent
handsomely on Donald Trump’s reelection, and he handed them the pen to write
whatever executive orders gave them carte blanche.
But hundreds of listening sessions
with Americans across political, income, and racial divides suggest a deeper
divide between how the people and the powerful are thinking about AI. When I
was a visiting scholar with Stanford’s Cyber Policy Center, conversations
started with your “P(doom)”—a shorthand term for the percent likelihood you
believe that AI will cause a catastrophic event—and went from there to
predictions of thresholds and timelines for superintelligence. AI symbolized
something revolutionary and without precedent, which it is.
Among tech leaders, AI
conversations focus almost entirely on x-risk, and the split is between the
accelerationists (mostly the investors profiting from AI who bet big on Trump
and federal preemption) and the AI safety crowd (mostly those actually creating
the technology who are demanding fail-safes, guardrails, and more). This clash
matters, and it is escalating rapidly, but the people fighting it often fail to
understand this as one major battle in the larger war over the future of
AI.
Back home, Americans experience AI
as an accelerant of realities already present in their lives that neither party
has had the courage to face. Most people are experiencing AI as
an adrenaline shot to automation trends that have been displacing human jobs
for years. AI makes it even harder for parents to protect kids from dopamine
and screen addiction, depression, and worse. AI is accelerating climate change,
as well as the erosion of privacy and civil liberties. It has flooded what’s
left of the public square with polarizing disinformation and slop.
AI is supercharging speculative
market swings and private equity’s power to squeeze even more wealth from Main
Street and the middle class. And America’s system of legalized corruption,
known as campaign fundraising, makes it easy to translate this unprecedented
concentration of wealth and data into political power.
I sincerely appreciate those inside
the tech world who have spent years calling for federal action on x-risk. But
their observation bias has made it harder to build the broad coalition needed
to deliver those reforms. They elevate the very real risk they understand
better than the rest of us, often to the exclusion of risks the rest of us know
from lived experience—risks that could destroy our livelihood and lives, our
agency and democracy, our planet, and even our most intimate spaces.
It is not lost on most Americans
that the powerful people who have reaped the largest benefits from these trends
are also the most evangelical about accelerating them. If those trends have
increased your wealth or power, the logical conclusion would be to let AI rip.
If you have been part of the 90 percent of Americans who have spent decades
watching the gains go somewhere else, few things sound more out of touch with
reality than these avatars of AI optimism wanting to pour fuel on the fire.
Voters are not asking for anything
unreasonable. They want four basic things.
First, most Americans have the
common sense to oppose anyone having the power to build a machine for private
profit that creates the risk of human extinction. Nobody elected these
companies, and they should not control the fate of humanity.
This month, the chief scientist of
OpenAI wrote that “no one is prepared” for what is coming; the CEO of
Anthropic called on the industry to slow itself down; the CEO of OpenAI agreed;
and more than a thousand employees of the AI companies signed a petition asking
the federal government to help pace the frontier. Congress should treat this
like the emergency it is and set national safety standards this month—or be
voted out in November to make room for leaders who will. And the tech leaders
expressing such concern with one hand should stop using the other to defeat the
candidates demanding such reforms.
The second request from the
American people is knowing that products they buy are reasonably safe for them,
their kids, and their aging parents. When folks around here buy a lawnmower,
they can reasonably assume it will not blow up or chop off their leg because of
regulations and liability laws that incentivize the company to produce a safe
product. If we can expect that from a mower, common sense suggests we should
have at least that for the most powerful tool ever put in human hands.
Sometimes the exception proves the rule: Washington exempted the big internet
and social media companies from that kind of liability, and families and entire
countries are still paying for that choice. Congress must move to set national
guardrails and liability standards that create the right incentives inside the
companies to value safety.
Third, people want a say in the
decisions that shape their lives. They want agency over how and when technology
enters their lives, whether that is protecting privacy or preventing data
centers and transmission lines. Americans have a well-founded animosity toward unchecked power and a 250-year commitment to the inalienable right of being
governed only with the consent of the governed.
Fourth, and perhaps most
challenging, is that Americans want jobs, they want their kids to have jobs,
and they want those jobs to pay enough to support a family without going into
debt. For two generations, automation, globalization, and private-equity speculation—blessed
by leaders in both parties—have shrunk the middle class in size and in its
share of national income. If paychecks had grown with the economy since the
late 1970s the way they did in the decades after World War II, the typical
full-time worker would be earning about $32,000 more a year. Where AI
accelerationists say, “Don’t worry, automation has been around for
years,” the American people are shouting back: We are worried precisely
because we have seen what automation has been doing to our jobs and paychecks
for years.
These midterms are the first
American election in which AI is a defining issue. In polls of competitive
House districts this spring, a new federal agency to monitor AI drew 75 to 83
percent support, including two-thirds to 85 percent of Republicans. President
Trump and this Congress chose the opposite direction, siding with the
accelerationists in almost cartoon-villain fashion.
While the American people are
demanding protections from AI, Trump and his allies in Congress made three
audacious attempts to give the AI companies a decade of carte blanche. After
several tech titans bankrolled Trump’s reelection, their first priority was to
roll back the modest but real safety rules put in place by the Biden administration. That was just the beginning. Their “Holy Grail” was
called “federal preemption,” and it would shield AI companies from
state liability and regulation for 10 years. After both red and blue states had
passed laws protecting kids, Trump’s proposal would make it illegal for states
to regulate AI—without providing any national standards to replace these
important efforts.
The Senate stripped the ban out, 99–1, with the encouragement of attorneys general from 40 states. Undeterred,
Trump delivered Big Tech’s return on political investment by executive order: the Justice Department directed to sue any state that puts up guardrails and
liability related to AI, and the federal funds appropriated to expand rural
broadband held hostage until states back down. Congress should not preempt
states from writing their own rules on AI at the industry’s request. Let
Virginia protect Virginians.
I am running for office in an area
that has experienced generational economic setbacks before. When I first
represented Southside Virginia in Congress, the smokestacks of Dan River Mills
had just come down. For a century, that mill wove the sheets America slept
on—14,000 jobs at its peak, in a city of 40,000. By 2006, the last of the work
had moved overseas. Families lost paychecks and pensions, young people left to
find jobs, and a proud city was told its best days were behind it.
Folks in Southside sometimes refer
to AI as “Nafta on steroids.” More accurately: AI is the new
globalization, and federal preemption is the new Nafta. It is the test people
will remember for a generation about which leaders reached for water and which
for gas when their town was on fire. Workers in the 1990s did not expect
President Clinton or the Republicans in Congress to stop globalization, but
they have never forgiven them for accelerating it.
People around here have never
forgotten the smug certainty of those who stood to profit, personally and
politically, from those votes, as they claimed that every factory worker would
become a programmer and China would become a liberal democracy. It is that same
smug confidence from a similar chorus of naïve geniuses that echoes through
today’s evangelists of acceleration. Imagine how it lands with people I know
who have a three-ring binder of certificates from every retraining program they
were told to complete, while each new job paid less and lasted fewer years than
the one before.
This is why the American people are
not just demanding safety. They are demanding a seat at the table. They do not
believe that the people who got rich off decisions that made them poor should
be trusted to set the rules for something likely to change every aspect of
life. We need immediate action to prevent catastrophic, personal, and localized
harms. These will prove to be defining issues of this election: regulations,
liability, and data centers. The next election will be defined by which
presidential candidate has the most serious answer to the much harder questions
of how we protect human agency and human employment in the era of AI.
The
politics of AI are not yet coded as red or blue. For all that is new about
AI, it elevates principles that are as timeless as they are universal—that
legitimacy only comes from the consent of the governed or the inalienable
rights to life, liberty, and the pursuit of happiness. The people I meet—the
ones with the three-ring binders—have accumulated far more wisdom than the
wealth amassed by the monotone disrupters. We are prepared to know exactly what is coming, and this time
they intend to be at the table when the decisions get made. In November, they
will be.
