When AI becomes the hacker: Cybersecurity enters a new era

Another weekend conversation, and that too the continuation of GISEC 2026 CISO community meeting, the context was: There was a time when hackers relied on hoodies, energy drinks and an unhealthy relationship with command-line interfaces. Today, they have something far more dangerous, nothing other than “artificial intelligence”.

Reminds of a story that sounds less like cybersecurity news and more like a nature documentary narrated by David Attenborough after three cups of strong coffee, a team of independent security researchers reportedly used Anthropic's Claude AI to help identify and exploit vulnerabilities within OpenAI's internal code environment. In other words, one artificial intelligence was recruited to help expose weaknesses in the ecosystem of another.

If that sounds alarming, imagine watching a king cobra casually swallowing a venomous viper. The spectators are stunned not because snakes eat snakes, but because the predator appears unfazed by the poison.

Welcome to cybersecurity in 2026.

For years, technology companies have proudly showcased AI as humanity's newest digital ally. AI would write reports, summarize meetings, create artwork and answer life's difficult questions such as whether pineapple belongs on pizza. Yet beneath the glossy marketing brochures lies a more complicated reality, which is “AI can be both the lock and the lockpick”.

The recent incident demonstrates precisely that.

Researchers, acting within authorized bug-hunting activities, allegedly leveraged Claude's analytical capabilities to identify pathways into OpenAI's systems. While there is no indication that malicious actors were involved, the event highlights a rapidly emerging trend. AI is no longer just the target of cyberattacks. It is becoming an active participant in them.

Think of it as hiring Sherlock Holmes to investigate Professor Moriarty, only to discover Sherlock can read a million pages of documentation before breakfast.

The analogy of the king cobra and the viper is particularly fitting. Traditionally, cybersecurity assumed that dangerous tools were difficult to access and required specialized skills. However, modern AI systems possess extraordinary abilities to analyse code, identify weaknesses, automate testing and even suggest exploitation methods.

When one powerful AI model is directed toward another platform, the result resembles nature's apex predators turning their attention toward each other instead of smaller prey.

The concerning part is not that researchers found vulnerabilities. That is exactly what security research is supposed to accomplish. The truly fascinating aspect is the speed and scale now possible.

A vulnerability assessment that might have taken weeks of manual effort can potentially be accelerated by AI assistants that never get tired, never forget syntax and never complain about staying late for penetration-testing exercises.

Cybersecurity teams are witnessing a profound shift. The old arms race involved hackers versus defenders. The new era increasingly resembles AI versus AI.

On one side, attackers use AI to discover, automate and scale offensive operations. On the other, defenders deploy AI for threat hunting, anomaly detection and incident response. The battlefield is becoming less about humans pressing keyboards and more about intelligent systems competing at machine speed.

In effect, the jungle has become autonomous.

This raises a critical question for organizations worldwide,that “what if AI can help uncover vulnerabilities in highly sophisticated AI companies, but what chance does the average enterprise have?”

The answer is both reassuring and unsettling.

Reassuring because responsible researchers continue to expose weaknesses before criminals can exploit them. Unsettling because every breakthrough in defensive AI simultaneously becomes a learning opportunity for offensive AI.

The cobra grows stronger. So does the viper.

Fortunately, the latest incident is also a reminder of why transparent security testing, bug bounty programs and independent research remain essential. Just as nature depends on predators to maintain balance within ecosystems, cyberspace depends on ethical researchers willing to challenge assumptions and stress-test even the most advanced technologies.

The lesson is simple that “AI is no longer just a tool. It is becoming an actor”.

And when a king cobra starts eating venomous vipers without breaking a sweat, the rest of the jungle should probably pay attention.

Because in the age of artificial intelligence, the question is no longer whether machines can help hack systems.

It is whether the next hacker will even be human.

We are in conversation more cybersecurity experts in our community, stay tuned…