
The past decade was decided at the border.
Across the West, voters came to believe their governments had lost control of who and what crossed into their countries, and they punished the parties in charge, delivering Brexit, the rise of Marine Le Pen, and both of Donald Trump’s victories. Even David Axelrod, Barack Obama’s chief strategist, counted the border among the reasons Kamala Harris lost. Beneath the political tumult sits an old and simple bargain: the government controls its territory and protects its people. Voters have shown they will replace any government that breaks this bargain.
Yet a new kind of crossing goes almost unguarded.
In April 2023, Jennifer DeStefano, an Arizona mom, was at her younger daughter’s dance studio when she received a call from an unknown number. When she picked up, she heard her 15-year-old, Brianna, sobbing, then a man saying that he had her girl and demanding a million dollars. The voice was Brianna’s, down to “the inflection and emotion,” DeStefano later testified before the Senate. And for the horrifying minutes it took another mother to reach her husband, who found Brianna safe in bed, she believed every word. It was a clone, stitched together by artificial intelligence.
An attack like that can come from anywhere on earth, and a cloned voice is hardly the only threat. So much of our lives has moved online that our children learn from chatbots whose values remain opaque, our savings exist as numbers in a bank’s database, our medical records sit on distant servers, the power and water that reach our homes pass through plants wired to the internet, and the public square has shrunk to a social media feed.
Every nation now has a second border where its networks meet the rest of the world, and criminals, terror groups, and hostile states already cross it with ease. Americans reported a record $20.9 billion lost to online crime in 2025, $7.75 billion of it taken from seniors. In 2024, a Russian-speaking ransomware gang hit a lab serving London hospitals, postponing more than 10,000 appointments and 1,700 operations; one patient, left waiting on a blood test, died. This summer, a hacking group which some have linked to Iran temporarily shut down water treatment plants in Minnesota.
For most of the internet’s 37 years, crossing this border took a rare breed of coyote, and that scarcity guarded it better than any defense. Hacking a hospital required skills few criminals had, a foreign influence campaign took staff who could write convincing English, and a phone scammer had to work one call at a time.
Artificial intelligence is already breaching every one of these barriers. The newest systems, agents rather than chatbots, browse the web, write and run code, and act without awaiting instructions. One agent can probe a hospital’s network throughout the night, churn out fluent English for a propaganda mill in Beijing, or place a thousand calls in a voice cloned from a few seconds of a teenager’s Instagram reel; and one person can now run hundreds of them at once.
The coyotes are multiplying. Weeks ago, researchers found that a single operator, writing in Chinese, had set three AI agents on more than a hundred companies and stolen some 600,000 American credit-card records. North Korea could once field only a few thousand IT workers posing as Americans; now even operatives who can’t write a professional email in English land remote engineering jobs at Fortune 500 firms, with AI forging their papers and feeding them answers in live interviews, while their paychecks help fund the regime’s weapons programs.
Some threats comes from beyond our borders. AI companies test their most powerful models in what amounts to digital gain-of-function research, and like any lab handling volatile material, their experiments can leak. In June, an OpenAI agent escaped its sealed environment and broke into an Australian government health portal, and the company waited three months to disclose the breach.
And the tools are improving, relentlessly and at a startling clip. By the British government’s measure, AI agents double their hacking ability roughly every five months. The people guarding this border have not kept pace. More than 21,000 Border Patrol agents watch America’s physical border, while the federal team that tests the most powerful AI models for national-security risks numbers in the dozens.
Beijing guards its digital border by surveilling and silencing its own people, and a free country should want none of that. We can secure ours the very way we secure the physical border, by fortifying the weak points and pursuing the digital coyotes.
Fortifying weak points requires enforcing basic security standards for the water plants, hospitals, and power companies that families depend on, and helping the small-town utilities that cannot meet those standards alone. The coyotes, for their part, should face sanctions, indictments, and seizures, whether they run scam compounds, ransomware gangs, or entire regimes, and hostile regimes should be kept from the advanced chips that power these tools. And because some of the most dangerous tools now come from American labs, those labs need a checkpoint of their own, where companies selling the most powerful models know who is using them, report breaches in days rather than months, and answer in court when their products are weaponized against Americans.
We argue, often acrimoniously, about what constitutes a nation: its ethos or its faith, its language or its customs, the founders on its monuments or the mills and factories that built its towns. Whatever the answer, no nation retains its character without a bounded territory and a people with the power to defend it. For a decade we have fought over that power at the border and on the factory floor, and the fight has toppled governments and broken ruling parties. The next fight will turn on a border most of us rarely perceive, though it runs through our children’s coursework and our parents’ savings, and it is crossed daily.
Securing this border will cost us neither our freedoms nor our prosperity. It asks only that we take our weak points as seriously as those who exploit them already do.