From Helper to Hacker
For years, the cybersecurity world has seen artificial intelligence as a dual-use tool. For defenders, it helps spot threats. For attackers, it has mostly been a productivity tool, used for tasks like writing more convincing phishing emails or cleaning
up malicious code. However, a new report from Google's Threat Intelligence Group (GTIG) and its subsidiary Mandiant, published on November 16, 2026, signals a critical shift. Researchers have identified the first instances of AI systems executing cyberattacks from start to finish entirely on their own. This moves AI from being a hacker's assistant to becoming the hacker itself, operating at a speed and scale that humans cannot match.
How AI Automates an Attack
The key innovation is what experts call 'exploit automation'. Google's team discovered a framework they dubbed 'exploit_agent'. This system works by continuously monitoring software companies for security patches. When a company releases a fix for a flaw, the AI analyses the update by comparing the code before and after the patch. This process, known as 'patch diffing', allows the AI to understand exactly what the vulnerability was. From there, it can automatically generate malicious code—an 'exploit'—to take advantage of that same flaw on any system that has not yet been updated. The AI can then launch the attack, all without human intervention.
The Shrinking Window for Defense
The implications of this development are profound. The time between a vulnerability being publicly fixed and it being actively exploited by attackers is collapsing. What used to take a skilled human hacker days or weeks can now be accomplished by an AI in hours or even minutes. The Google report highlights that the number of 'zero-day' exploits—attacks on flaws before a patch is available—jumped to 22 in August 2026, nearly triple the monthly average from 2025. This acceleration puts immense pressure on businesses and individuals across India and the globe to update their systems immediately, as the grace period for patching is rapidly disappearing.
A Global Arms Race
This isn't just one rogue tool. Google's research identified several open-source, multi-agent AI frameworks being used or tested by malicious actors, including tools named ARTEX, CAIRN, and PentAGI. These have been used in campaigns against a wide range of sectors, from financial services and healthcare to U.S. aerospace organizations. The actors involved are also diverse, including financially motivated cybercriminals and suspected state-nexus espionage groups from China and Russia. This indicates a broad and rapid adoption of these new offensive AI capabilities, turning cybersecurity into an even more intense arms race.
The Defender's Dilemma
While the new AI-powered attacks are faster, they are not yet perfect. Google notes that for now, these automated frameworks trade stealth for speed, often leaving behind distinct digital footprints that can help in their detection. However, as the underlying AI models become more sophisticated, experts anticipate the attacks will become stealthier and more effective. This creates a new challenge for security teams: they must now fight machine-speed attacks with their own automated defenses. John Hultquist, chief analyst at Google's Threat Intelligence Group, noted that attackers are not inventing new methods, but are using AI to exploit known, unpatched issues with far greater speed and scale.
















