The conversation around AI and cybersecurity has often been speculative. But according to the latest findings from Google's Threat Intelligence Group, the future is here. Attackers are no longer just using AI; they are deploying it autonomously.
A Paradigm Shift in Cyberattacks
The latest
report from Google's Threat Intelligence Group (GTIG) and its subsidiary Mandiant, released in November 2026, marks a significant turning point in the cybersecurity landscape. For the first time, researchers have observed threat actors deploying fully autonomous AI frameworks to discover vulnerabilities, create exploits, and execute attacks without direct human intervention. This represents what Google calls a 'paradigm shift'—a move from AI-assisted hacking to intrusions occurring at machine speed. While the core attack methods haven't changed, as they still often rely on weaknesses like unpatched systems and stolen credentials, the speed and scale at which they can now be executed present a formidable new challenge for defenders.
The Rise of Autonomous Agents
Hackers have moved past simply asking a chatbot to write malicious code. They are now using open-source, multi-agent frameworks with names like ARTEX, CAIRN, and PentAGI. These systems can chain together complex attack sequences, allowing different AI agents to collaborate and share information to achieve their goal. Google's report details a system dubbed 'exploit_agent', which was found on an exposed server. This framework was designed to automatically compare software code before and after a security patch was released. By analyzing the differences, it could identify the vulnerability the patch was meant to fix, then write and execute an attack to exploit it on systems that hadn't been updated yet. This capability dramatically shrinks the critical window between a patch being issued and it being reverse-engineered for malicious purposes.
Speed and Scale Over Stealth
One of the key findings is that these AI-driven attacks currently sacrifice stealth for speed and scale. An autonomous system can scan for vulnerabilities and launch exploits against thousands of targets far faster than a human operator. For example, Google observed the CAIRN framework being used in attacks against the US aerospace sector, and a customized version of PentAGI scanning organizations across Asia. This approach has proven effective for both financially motivated cybercriminals and state-sponsored espionage groups, with actors linked to China and Russia already experimenting with or deploying these tools. The report also highlights a new monetisation strategy called 'LLMjacking', where attackers compromise cloud accounts to steal expensive AI computing resources, sometimes racking up fraudulent charges exceeding $100,000 in a single day, and then resell that access on underground markets.
Implications for Indian Businesses
For India's rapidly digitizing economy, these developments are a critical warning. As businesses increasingly adopt cloud infrastructure and AI tools, their attack surface expands. The speed at which AI can weaponize a newly disclosed vulnerability means that traditional, manual approaches to security are no longer sufficient. The time it takes to patch systems is now more critical than ever. According to Google's findings, exploitation remains the leading initial infection vector for the sixth consecutive year. The rise of AI-powered phishing and social engineering also poses a significant threat, as these tools can now craft highly convincing, personalized messages in any language, bypassing traditional email filters and tricking employees. This automation lowers the barrier to entry, allowing less skilled actors to launch sophisticated campaigns.
















