Fighting Fire with Fire: Can Good AI Defeat Bad AI?

Bill Gates Warns of an AI Cybercrime Wave. Maybe Artificial Intelligence Is Also Our Best Defense.

Part of the Signals From the Future collection — observations on AI society, synthetic reality, digital culture, Chatrodamus Prophecies, emotional technology, and humanity’s increasingly complicated relationship with machines.

Bill Gates is sounding the alarm again.

This time, he’s warning that artificial intelligence could unleash a wave of sophisticated cyberattacks within the next year, threatening banks, businesses, financial markets, and potentially the systems we depend on every day.

According to a Fox News report about his appearance on CNN’s State of the Union, Gates warned that criminals with access to powerful AI models and few safeguards could deliver what he called “a lot of negative surprises.”

That’s a polite way of saying we might be in for one hell of a mess.

But while everyone is busy worrying about what the bad guys might do with artificial intelligence, I have a question.

Why aren’t we talking more about using artificial intelligence to stop them?

In other words, why not fight fire with fire?

Not by unleashing some rogue AI vigilante to hunt down hackers and blow up their computers, but by developing defensive AI systems capable of detecting, blocking, and containing cyberattacks faster than human beings ever could.

Because if criminals are going to use AI as a weapon, we’d better make damn sure the good guys have better shields.

---

The Cybercriminals Are Getting Smarter Weapons

Cybercrime isn’t new.

We’ve had computer viruses, identity theft, phishing scams, ransomware, and financial fraud for decades.

What’s changing is the technology available to the people committing those crimes.

In the past, a cybercriminal needed considerable technical knowledge to develop sophisticated malicious software or conduct complex attacks.

Artificial intelligence can potentially lower some of those barriers.

It can help criminals write more convincing phishing emails, impersonate executives, automate parts of their operations, and analyze stolen information at tremendous speed.

Imagine receiving a telephone call from your bank manager.

You recognize his voice.

He knows your name.

He even sounds concerned about suspicious activity on your account.

There’s just one problem.

It isn’t your bank manager. It’s an AI-generated impersonation designed to steal your money.

Now multiply that possibility by thousands, or even millions, of potential victims.

That’s the kind of future we’re approaching.

And while AI doesn’t magically make every criminal a master hacker, it can make certain kinds of fraud cheaper, faster, and easier to scale.

---

Bill Gates Isn’t the Only One Concerned

The Fox News article cites CrowdStrike’s 2026 Global Threat Report, which reports an 89% year-over-year increase in attacks attributed to AI-enabled adversaries during 2025.

It also reports an average lateral movement time of 29 minutes—the time attackers take to move from an initial compromised system toward other systems in a network.

Those are concerning figures, although they measure different aspects of cyber threats and don’t mean that every attack was carried out autonomously by AI.

Think about what 29 minutes means.

You could sit down for breakfast, drink a cup of coffee, read the headlines, and discover that someone had already moved through critical parts of your company’s computer network.

And the attackers aren’t necessarily working alone.

They may be supported by software that can analyze enormous quantities of information and assist with tasks far faster than a human operator.

But here’s the part that interests me.

If AI can help criminals find weaknesses, why can’t AI help defenders find those weaknesses first?

---

Fighting Fire with Fire

I spent thirteen months in Vietnam as a United States Marine.

One thing I learned was that every offensive weapon eventually inspires some form of defensive response.

Mines led to mine detectors.

Aircraft led to radar and antiaircraft defenses.

Missiles led to missile defense systems.

Electronic warfare led to electronic countermeasures.

The basic principle hasn’t changed.

When somebody develops a new way to attack you, you develop a new way to defend yourself.

Artificial intelligence is simply the latest battlefield.

Only this time, the battlefield isn’t necessarily a jungle, a desert, or an ocean.

It could be the computer network that keeps your local hospital operating.

It could be the electrical grid.

It could be your bank.

Or it could be the smartphone sitting in your pocket.

And unlike conventional warfare, an AI-assisted cyberattack might unfold at a speed that leaves human defenders struggling to keep up.

That’s why I believe AI-assisted defense is not merely an option. It’s becoming a necessity.

---

Meet the Good AI

Let’s imagine a sophisticated defensive AI protecting a major bank.

Every second, it analyzes network traffic, account activity, login attempts, software behavior, and thousands of other signals.

Most of what it sees is perfectly ordinary.

Customers checking balances.

Businesses transferring funds.

Employees doing their jobs.

Then something changes.

A computer begins accessing information it has never accessed before.

An employee account suddenly attempts to transfer money to an unfamiliar destination.

A suspicious program starts encrypting files.

A login appears from an unusual location, followed by activity inconsistent with the account holder’s normal behavior.

Individually, those events might have innocent explanations.

Together, they could indicate an attack.

A defensive AI could identify the pattern, flag the activity, restrict a compromised account, isolate an affected computer, and alert human security specialists.

Some of these capabilities already exist in modern cybersecurity products, using machine learning, automated rules, behavioral analysis, and increasingly AI-assisted investigation.

We’re not starting from scratch.

But future systems could become more capable of connecting the dots and recommending defensive actions in real time.

And that raises an interesting possibility.

The first system to recognize an AI-assisted attack might be another AI.

---

Call the C.O.P.S.!

We have police officers patrolling our streets, security guards protecting our businesses, and military personnel defending our country.

So who’s patrolling cyberspace?

I think it’s time to call the C.O.P.S.

Cybernetic Officers for Protection and Security.

That’s my name for the next generation of AI-powered cybersecurity defenders.

Imagine millions of digital officers patrolling computer networks around the clock, searching for suspicious activity, identifying cybercriminal behavior, and stopping attacks before serious damage occurs.

They don’t need coffee breaks.

They don’t collect overtime.

They don’t call in sick.

And they certainly don’t need doughnuts.

But unlike human police officers, these C.O.P.S. could potentially monitor enormous volumes of digital activity simultaneously, identifying patterns that even the best human cybersecurity specialists might miss.

Of course, there’s one important difference between my imaginary C.O.P.S. and the police officers walking a beat.

These digital cops shouldn’t have unlimited authority.

They need rules, supervision, accountability, and clearly defined limits on what they can do.

Think of them as digital guardians rather than judge, jury, and executioner.

Let them identify the threats, raise the alarms, and contain attacks within authorized systems.

Leave the really consequential decisions to human beings.

Because the last thing we need is a bunch of overzealous robot cops shooting first and asking questions later.

The bad guys have their AI. It’s time the good guys called the C.O.P.S.!

---

When Machines Fight Machines

Now things get interesting.

Suppose a criminal uses AI to probe a company’s network for weaknesses.

A defensive AI detects the suspicious activity and blocks it.

The attacker’s tools adjust their approach.

The defensive system identifies the change and strengthens its response.

Back and forth they go.

Attack.

Defense.

Adaptation.

Countermeasure.

All happening at machine speed.

Humans might supervise the engagement without being able to examine every individual action as it occurs.

We’ve seen versions of this contest before in computer security, where attackers and defenders continuously adapt their techniques.

But increasingly capable AI could accelerate the process.

And here’s what bothers me.

What happens when the machines start making decisions faster than we can understand them?

Cybersecurity isn’t the only field where artificial intelligence could make decisions faster than humans. In The Future of Air Travel, I examine how emerging technology could transform aviation. Whether we’re protecting a computer network or operating an aircraft, the underlying question is remarkably similar: how much control should we surrender to intelligent machines?

---

The Danger of Giving Good AI Too Much Power

There’s an old saying about fighting fire with fire.

Sometimes you end up burning down the whole damn neighborhood.

That’s the danger of giving defensive AI unlimited authority.

Imagine an AI protecting a hospital network.

It detects suspicious activity and decides to shut down communications to prevent an attack.

Unfortunately, those communications support legitimate medical operations.

The system may have stopped a cyberattack while creating another emergency.

Or imagine an AI incorrectly identifying an innocent computer network as the source of an attack and launching its own retaliatory operation.

Now we’ve crossed a dangerous line.

Defending your own systems is one thing.

Attacking somebody else’s systems is something entirely different.

Attribution in cyberspace is notoriously difficult. Criminals often route attacks through compromised computers belonging to innocent people and organizations.

A defensive AI that automatically strikes back could end up harming the wrong target.

That’s why I believe we need to distinguish between automated defense and autonomous retaliation.

I’m comfortable with an AI recognizing malicious activity, quarantining compromised systems, and alerting security personnel under carefully designed safeguards.

I’m considerably less comfortable with an AI independently deciding to attack another network.

Call me old-fashioned.

But before a computer starts a war, I’d like a human being to have something to say about it.

---

The Human Factor

Here’s another problem.

Artificial intelligence can be fooled.

It can misinterpret information, produce incorrect conclusions, and be manipulated by attackers.

A criminal might deliberately feed misleading data into a defensive system.

Or craft instructions designed to make an AI security assistant ignore warnings or reveal information it shouldn’t.

So merely putting an AI in charge doesn’t guarantee safety.

We still need cybersecurity professionals, independent testing, human oversight, carefully restricted permissions, and systems that can be shut down or overridden when necessary.

And if you want practical advice instead of predictions, the Cybersecurity and Infrastructure Security Agency (CISA) offers free guidance on protecting yourself online, including recognizing phishing scams, using multifactor authentication, creating strong passwords, and keeping your software updated. You don’t need your own personal C.O.P.S. patrol to start defending yourself against cybercriminals today.

We also need something less glamorous.

Basic computer security.

Strong authentication.

Regular software updates.

Reliable backups.

Employee training.

A company with lousy security practices doesn’t become invincible just because somebody installed an AI-powered dashboard.

That’s like putting a sophisticated alarm system on a house while leaving the front door wide open.

The question of how much responsibility we should hand over to machines goes far beyond cybersecurity. In my article The Last Human Job, I explore what might remain uniquely human when artificial intelligence becomes capable of performing more and more of our work. Perhaps one of our most important responsibilities will be deciding when the machines must stop and let humans take charge.

---

What About Ordinary People?

You don’t have to operate a bank or manage a corporate computer network to be affected.

Consider retirees.

Millions of older Americans depend on online banking, Social Security payments, retirement accounts, and electronic communications.

I’m one of them.

Living in the Philippines, I depend on technology to communicate, manage finances, and stay connected to the world.

An AI-assisted scammer doesn’t care whether you’re sitting in Manhattan, Missouri, or Mindoro.

If he can convince you to reveal your banking credentials or authorize a fraudulent payment, he’s happy.

Defensive AI might help banks identify suspicious transfers or detect unusual account activity.

But it won’t eliminate the need for common sense.

If somebody calls claiming to be your bank, hang up and call the institution using a number you independently know is legitimate.

If an urgent message demands money or sensitive information, verify it through another trusted channel.

And don’t assume a familiar voice or realistic video proves you’re speaking with the real person.

In the age of AI-generated impersonations, seeing and hearing may no longer be believing.

---

Are We Entering an AI Arms Race?

I think we are.

Not necessarily an arms race between countries, although that dimension certainly exists.

I’m talking about an escalating competition between those who use artificial intelligence to exploit systems and those who use it to protect them.

The criminals will search for vulnerabilities.

The defenders will develop better detection.

The criminals will change tactics.

The defenders will adapt.

And the cycle will continue.

There may never be a final victory.

Cybersecurity has always been an ongoing contest, and AI could make that contest faster and more complicated.

But I don’t accept the idea that the only possible response is to fear the technology.

A hammer can build a house or smash a window.

A knife can prepare dinner or injure someone.

Technology doesn’t become good or evil simply because it is powerful.

What matters is how people develop it, deploy it, control it, and remain accountable for its consequences.

This competition between offensive and defensive artificial intelligence is part of a much larger question I’ve been exploring in my post: How Much Authority Should We Give AI? what happens when the technologies we create become powerful enough to change society faster than we can adapt? The real challenge may not be inventing smarter machines. It may be remaining smart enough to control them.

---

The Real Question Isn’t Whether AI Will Be Used

Bill Gates is warning us that increasingly powerful AI could give cybercriminals new opportunities.

I think we should take that warning seriously.

But I also believe the answer cannot be to surrender the technological advantage to the people who intend to misuse it.

We need better defensive systems.

We need smarter cybersecurity professionals working alongside AI.

We need rules governing what automated systems can do, particularly when their actions could affect critical infrastructure or innocent third parties.

And we need to remember that artificial intelligence is a tool—not a substitute for human judgment and responsibility.

As a Marine, I learned that you don’t defeat an enemy simply by hoping he’ll stop shooting.

You improve your defenses.

You anticipate his next move.

And you make sure you’re prepared.

The same principle applies here.

If bad AI is coming for our computer networks, good AI had better be waiting at the gate.

But we’d better make damn sure we still control the gate.

Because the future of cybersecurity may not be humans fighting machines.

It may be machines fighting machines—with humanity caught in the middle.

---

---

---

---

Chatrodamus | Signals From the Future

Source note: This commentary was inspired by the October 2026 Fox News report supplied for this article, quoting Bill Gates’ CNN interview and citing CrowdStrike’s 2026 Global Threat Report. Gates’ forecast is a prediction, not an established outcome. The hypothetical scenarios and future possibilities discussed here are illustrative, not reports of actual incidents.