The Frontispiece of Leviathan, The Matter, Forme and Power of a Common Wealth Ecclesiasticall and Civil by Thomas Hobbes, published in 1651. The engraving is by Abraham Bosse.

...

Shane Croucher

By Shane Croucher

Senior Editor

0

Share

Newsweek is a Trust Project member

See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.


On Monday, Google’s Threat Intelligence Group reported a highly alarming first in our new age of artificial intelligence. Attackers had used an AI framework to find and exploit flaws in software without a human in the loop. Put simply, it was a fully automated cyber attack.

It comes amid a wave of highly disruptive—and allegedly coincidental—patch-driven outages across government and private business, from air traffic control systems, to IT service providers, to banks, and much more. A taste, perhaps, of the cyber terror to come.

The system Google flagged was called "exploit_agent." It had compared the code of popular web software before and after each security patch, worked out what vulnerability the fix had revealed, then wrote and launched an exploit all on its own.

Read More on Analysis

Stolen victim data sat on the same server. Its configuration notes were, tellingly, in Russian. The operator appeared focused on targeting French websites. Google’s companion post described an attacker that "doesn’t eat, sleep, or get tired."

The philosopher Thomas Hobbes, born centuries earlier, would have recognized this terrain.

In Leviathan, published in 1651, he argued that the anarchic state of nature is life without a common power able to keep everyone in check, a condition in which no one can be confident they won’t be attacked, so everyone is suspicious, armed—and strikes first.

That description fits the current global internet. Humanity has built a border-crossing arena of permanent competition in which attackers can operate beyond the reach of their victims’ governments, while ordinary citizens increasingly provide for their own security.

Artificial intelligence is now intensifying that imbalance. The internet may have laws in different jurisdictions, even if the technology is largely borderless. But it still has no common sheriff sitting above them.

Related Story

Paris Hilton: Taking Down My Deepfakes Isn’t Enough | Opinion

Paris Hilton: Taking Down My Deepfakes Isn’t Enough | Opinion

Hobbes rooted his argument in a grim equality. Even the weakest can kill the strongest, by stealth or by alliance. Cyberspace industrializes that insight.

Microsoft’s 2026 Digital Defense Report notes that defenders may guard thousands of systems while an attacker needs only one way in. "AI is increasingly helping find that path," the report says.

The median time from a flaw’s discovery in the wild to its weaponization is now well under 24 hours. Yet fixing critical exposed vulnerabilities can take companies 30 to 60 days.

AI widens that gap by making aggression much faster and cheaper. Microsoft stresses that fully autonomous attacks are not yet the norm. But complex work that once demanded rare expertise can increasingly be handed to clever software that produces instant results.

Whatever the cause, the pressure is mounting. Google counted 22 exploited zero-days in August alone, against an average of eight a month in 2025. The exploit_agent system didn’t even need those. It fed on fixes already published, preying on those too slow to patch.

John Hultquist, chief analyst at Google Threat Intelligence Group, warned that companies now face "enormous pressure to speed up their patch cycles in the near future, and may even have to make some hard choices to keep up."

Related Story

Explained: China’s ‘Singapore Washing’ at Heart of Nvidia Scandal

Explained: China’s ‘Singapore Washing’ at Heart of Nvidia Scandal

Hobbes had anticipated readers who found him too cynical in his perception of human nature.

Such skeptics intuit what he knew to be true when they arm themselves for a journey, lock their doors at night, and their chests at home, even knowing there are laws and armed officers to avenge any wrong.

Online, Americans lock their chests with password managers, two-factor codes, and credit freezes. Increasingly, that lock is the whole defense, and it may too become vulnerable to AI’s advances.

Local police can’t arrest a server abroad. Even if they do pin down an individual suspect, bureaucracy, borders, and geopolitics stand in their way. Meanwhile, hospitals, banks, and small businesses race against AI attackers on their own slow patch schedules.

Hobbes was interested in the social contract. To be fair, an international social contract exists, but it’s in fragments.

The Council of Europe’s Budapest Convention, the first international cybercrime treaty, now has 84 parties. The newer U.N. Convention against Cybercrime has drawn more than 90 signatures but only three ratifications, far short of the 40 it needs to take effect.

Hobbes put it bluntly: "Covenants, without the sword, are but words and of no strength to secure a man at all."

Related Story

Major Banks Go Down in America’s Everything-Is-Broken Era

Major Banks Go Down in America’s Everything-Is-Broken Era

Objectors might point out that criminal law applies online too. The International Committee of the Red Cross, for example, insists cyberspace is no "legal vacuum," even in wartime.

Last week, police in nine countries, including the U.S., seized 235 servers and 257 domains behind three malware operations. And AI arms defenders, too. Microsoft argues the same advances enable security teams to respond at machine speed.

All of which is true. But Hobbes’s state of nature is useful here as a map of where protection ends: at the edge of whoever can actually enforce.

Who is closest to the role of internet sheriff? Google found exploit_agent and, in the same announcement, pitched its AI defense service. That’s Hobbesian too. He said allegiance lasts only as long as the sovereign can protect, drifting to the sword that actually does.

But that also points to the real danger. Hobbes’s remedy was the Leviathan, a sovereign mighty enough to overawe everyone and entitled to vet books before publication.

A digital Leviathan able to identify every user and compel every platform would be the surveillance state perfected.

The U.N. treaty began as a Russian proposal, which hardly inspires confidence. Digital-rights advocates warn it opens cross-border surveillance channels without enforceable safeguards. The U.S. declined to sign it in Hanoi last year.

Nor will more rules alone be enough. As the cryptographer Bruce Schneier argued this month, capable AI will find hacks wherever rules leave loopholes.

"Technically, we need to understand how AIs hack systems, build systems that are resilient to hacking and build AIs that won’t act as genies," he wrote.

"Politically, we need to learn to govern and regulate at the speed of technology. Otherwise, market failures will overwhelm our ability to govern them."

The internet needs something tougher to achieve than a global Leviathan: enforceable rules among states that share democratic safeguards, cross-border takedowns that are routine rather than occasional, and defenses that run at attackers’ speed.

We embraced the internet as an escape from borders and constraints. We are discovering how much security depends on rules, institutions and the power to enforce them. Hobbes wouldn’t be surprised. He’d ask who holds the sword and on what terms.

View 0 Comments

Request Reprint & Licensing

Submit Correction

View Editorial & AI Guidelines