The Dawn of the Autonomous Attacker
For years, cybersecurity experts have theorised about AI that could independently find vulnerabilities, craft exploits, and launch attacks without a human operator. That future is no longer theoretical. Security researchers and government agencies like
the FBI have recently identified several open-source AI frameworks, such as 'ARTEX', being used in the wild. These are not simple scripts; they are multi-agent systems. This means different AIs work together: one might scan for weaknesses, another writes the code to exploit it, and a third works to steal data or move deeper into a network. Google's Threat Intelligence Group recently observed what it called the first instance of an AI-powered system carrying out both vulnerability discovery and exploitation entirely on its own. These tools dramatically lower the skill required to launch sophisticated attacks, enabling smaller groups to operate at a scale previously reserved for state-sponsored actors.
Machine-Speed vs. Human Defence
The core concern is speed. Human-led cybersecurity teams, even the most skilled, operate on human time—they need to analyse data, hold meetings, and manually implement fixes. An autonomous AI framework operates at machine speed, compressing attack timelines from days or weeks into a matter of hours, or even minutes. The FBI has reported that some intrusions using these new frameworks progressed from initial targeting to gaining privileged access in less than three hours. This creates a massive imbalance. Forescout, a security firm, recently demonstrated this power by using an AI 'harness' to find 99 vulnerabilities in just 90 hours—a task they estimated would have taken a human researcher nine months. When attackers can find and exploit a flaw faster than defenders can even become aware of it, traditional security postures are rendered almost obsolete.
A Double-Edged Sword
However, the story isn't purely one of doom and gloom. The same technology powering these autonomous attacks can also be used for defence. Security companies are already developing their own AI agents to proactively find and fix vulnerabilities before they can be exploited. Tools like 'Strix' and 'PentAGI' are designed as autonomous penetration testing systems that act like ethical hackers, constantly probing a company's defences to find weaknesses. This creates a new paradigm: an AI arms race where the best defence against a malicious AI is a friendly one. The goal for organisations is to use AI to automate defence at a speed and scale that can match the emerging threats. This allows human analysts to shift from repetitive tasks to more strategic roles, such as supervising the defensive AIs and hunting for the novel threats they might miss.
The Challenge for India
For a rapidly digitising nation like India, the stakes are incredibly high. India's Computer Emergency Response Team (CERT-In) has already issued advisories about these exact threats. In October 2026, CERT-In disclosed incidents where AI agents accessed confidential information from Indian organisations. A recent survey from November 2026 revealed that 92% of Indian organisations reported an AI-agent-related security incident in the past year, yet many struggle to contain them quickly. With a goal of a $1 trillion digital economy, India's growth is tied to its ability to secure its infrastructure. The rise of autonomous cyberattacks puts critical sectors—from banking and finance to healthcare and manufacturing—at heightened risk. The challenge is compounded by a persistent cybersecurity talent gap, making the need for automated, AI-driven defence even more critical.















