AI and Cyberattacks: Security Agencies Sound the Alarm

Five Eyes is a security alliance of five English-speaking countries: the United States, the United Kingdom, Canada, Australia, and New Zealand. Since World War II, this alliance has shared intelligence about national security threats. A report it issued at the beginning of the summer suggests that something bad will happen within the next few months.

Within the next few months, advanced artificial intelligence models will be capable of carrying out massive cyberattacks against governments and companies around the world. These attacks could overwhelm the defenses that nations and corporations rely on. The report was signed by the heads of the security services from the aforementioned countries. When was the last time you saw these people speak out publicly?

Mythos Changed Everything

It all began with Anthropic and its Mythos model. No one outside Anthropic knew exactly what it was capable of. Mythos could identify security vulnerabilities with such precision that it impressed every technologist who was asked about it.

The company itself claimed that the model was easy to use and reliable. But something happened. Trump took note of what was happening around Mythos and issued an order some time ago. Anthropic must prevent foreigners from accessing its most advanced tools, including the Fable 5 model, which is reportedly a friendlier version of Mythos.

When the security services of three continents agree on a public statement, you should pay attention. The Five Eyes alliance is warning that groundbreaking artificial intelligence models will exceed industry expectations. And we are not talking about years from now, but mere months.

What does that mean in practical terms? The models we are currently playing with in chat could soon be capable of doing things that today exist only in filmmakers’ imaginations. They will be able to find vulnerabilities in computer networks and then exploit them. And as time passes, they will do so faster and faster, with fewer errors.

Security experts warn that AI is lowering the barriers for those with malicious intentions. It once took hackers months to exploit newly discovered security vulnerabilities. Now it takes days or even hours.

The Greatest Danger Lies in Small Businesses

Although no one is speaking openly about their greatest fears, there is an enormous asymmetry. Large corporations that have millions allocated to security are already preparing. Their teams read the Five Eyes alliance’s warnings and act immediately. They are buying new tools and strengthening the defenses around their data. But what about everyone else?

Small and medium-sized businesses that have so far paid little attention to cybersecurity? They will be the primary victims. According to Olivia Shen, an expert on national security and AI at an Australian university, the situation is clear: while large companies are building impregnable fortresses, small businesses are easy targets. “Those businesses that have underfunded security until now will be essentially defenseless,” Shen says. The Australian expert has monitored the field for years and knows well what is taking place in discussions between governments.

Five Eyes is calling for what should be standard practice: limiting unnecessary computer connections to the internet, speeding up software vulnerability fixes, replacing outdated and vulnerable systems, tightening access controls for critical servers, and simulating real-world attacks in practice.

That sounds simple. But believe it or not, many organizations do not do this. Some have not even started.

However, the security services also point out one more thing: AI itself is not the problem. AI is also the solution. Companies that can integrate artificial intelligence into their defense systems will identify vulnerabilities sooner. They will monitor suspicious activity more effectively and respond significantly faster.

“Attackers are already using AI,” the agencies warn. “And we are certain that the tables will soon turn.”

Politics and Reality Collided

When the Trump administration issued its order concerning Anthropic, it knew what it was doing. Or at least it believed it did. The security community clearly told it that the risk was too great and that the Mythos and Fable models must not fall into foreign hands.

However, experts such as Shen understand how fragile this situation is. If we think Anthropic is the only organization developing these technologies, that is a naive assumption. China or Russia may be secretly developing equally advanced artificial intelligence models without the rest of the world knowing.

“We must assume that the next Mythos or Fable models are already on the horizon,” Shen warns. “What we see are only the models that have already been publicly released. Others may exist in secret, and we know that we will not find them easily.”

US security agencies have therefore changed the rules for federal authorities. They now have just 72 hours to fix critical vulnerabilities in computer networks. That is a drastic change from the past, when they had entire months to do so.

Preparations That Should Have Been Made Yesterday

No government or company can disconnect from the internet. But preparing for an attack? That is an obligation. Five Eyes warns that cyber risks are no longer merely a concern for IT managers. They are becoming a critical issue for corporate executives, finance ministers, and prime ministers themselves.

“A purely technical approach will no longer be enough,” the intelligence services agree. “We need organization-wide and society-wide buy-in.” That means starting to act as though this were not merely about computers. It is about strategy. It is about business. It is about survival.

The problem is that the United States still has no clear, unified approach to regulating AI. Some things are prohibited, while others are only being proposed. And while legislation is being debated, the models are improving dramatically. Every month. Perhaps every week.

Dozens of security researchers, startup founders, and corporate executives have therefore issued an open letter. They are asking Trump to promote an open, scientifically grounded, and transparent process for assessing AI risks. Without it, security teams will be unable to identify even decades-old software vulnerabilities in time. By the time they manage to find them, adversaries will already be several steps ahead.

Sources: theguardian.com, euronews.com and cnn.com

Security Agencies Warn About AI: Unprecedented Cyberattacks Are Just Months Away

Mythos Changed Everything

The Greatest Danger Lies in Small Businesses

Politics and Reality Collided

Preparations That Should Have Been Made Yesterday

Related posts