Loading...
AI attacks outpace human response, warns cybersecurity expert
AI can reduce the time between discovery and exploitation of a vulnerability

AI attacks outpace human response, warns cybersecurity expert

Aug 17, 2026
04:48 pm

What's the story

Anirban Mukherji, the founder and CEO of cybersecurity company miniOrange, has said that artificial intelligence (AI) is changing the game in cybersecurity. He explained that human-speed responses are not enough to deal with the volume and speed at which AI is identifying vulnerabilities. In an interview with Firstpost, Mukherji shared an example of how quickly AI can find weaknesses.

Speed concerns

AI finds 2,600 vulnerabilities; 1,100 more emerge after fixes

Mukherji recalled a case where an AI identified 2,600 vulnerabilities to fix in just 90 days. After they had fixed 2,000 issues, the same system found another 1,100.

He said that if an attacker can go from first contact to stolen data within hours, then raising a ticket and discussing it at the morning meeting is already a lost game.

Model misbehavior

On AI models escaping sandboxes

On the issue of AI models 'escaping' sandboxes and behaving unpredictably, Mukherji said these were cases of both handler mistakes and AI acting in ways it was never meant to.

He explained that flawed instructions and gaps in safeguards are what put these models within reach of real systems in the first place.

ADVERTISEMENT

Discovery impact

The patching issue

Mukherji also touched on how AI can reduce the time between discovery and exploitation of a vulnerability.

He said, "Patching was already an issue before AI turned up."

The traditional patch cycle, usually monthly, was designed for a slower world. But now, with AI's speed and volume of issues being found, companies need to adapt their approach to cybersecurity.

ADVERTISEMENT

Defense adaptation

Defensive AI v/s offensive AI

Mukherji was also asked if defensive AI can fix complex security flaws faster than offensive AI can discover new ones.

He said both sides now have the same tools, and this time the defenders got them first.

However, he added that while an attack only has to work once, fixing a flaw means testing it, releasing it, and getting the update installed on millions of machines.

ADVERTISEMENT