AI Tools Flood Dark Web as Cybercrime Markets Surge 2700% in Four Months

AI Tools Transform Underground Cybercrime Markets

AI tools have flooded into underground cybercrime markets at an alarming rate, with advertisements surging from fewer than 50 per month in late 2025 to more than 1,400 by February 2026—a 2700% increase in just four months. According to research from the Halcyon Ransomware Research Centre1

1

, this explosive growth signals a fundamental shift in how cybercrime operates on the dark web. The barrier to entry has effectively vanished, with sellers offering access to ChatGPT Plus accounts for as little as $0.10 and jailbreak prompts designed to strip safety guardrails from mainstream models for just $0.322

2

. Halcyon analyzed nearly 4,000 posts across 77 Telegram channels, 20 dark web forums, and five specialized underground markets between October 2025 and May 20261

1

. What researchers discovered reveals an ecosystem that has achieved commercial maturity faster than anyone anticipated.

Cybercriminals Adopt Software-as-a-Service Business Models

Cybercriminals have completely commercialized the sale and creation of AI tools, mirroring legitimate software-as-a-service subscription models and online storefronts. Some sellers now promote AI tools tested on traditional dark web forums through easier-to-manage storefronts hosted on Telegram1

1

. The commercialization extends to pricing strategies that would look familiar to any tech startup. Roughly 40% of observed messages advertise free tools, creating a freemium funnel that mirrors legitimate software companies2

2

. "It shouldn't have been shocking to me," said Cynthia Kaiser, head of Halcyon's Ransomware Research Center and former deputy director of FBI Cyber. "But what we should take away from this is also how much cybercriminals have learned throughout the years about how to sell and package—and they applied this really quickly to the AI models"1

1

. This rapid adoption demonstrates that the AI cybercrime economy assembled tiered subscription pricing, automated distribution, multi-platform redundancy, and customer acquisition structures at the velocity of the legitimate AI industry—structures that took narcotics and exploit-kit markets years to build2

2

.

Four Categories of AI-Enabled Cybercrime Tools

The market for AI tools in underground cybercrime markets is organized around four distinct categories. First, jailbroken AI services and stolen accounts allow users to bypass safety guardrails from mainstream models for as little as $101

1

. Second, AI-enabled identity fraud and phishing tools make impersonating real people quick and easy, with social engineering now supercharged by artificial intelligence. Researchers observed a Russian-language vendor advertising a self-sufficient AI call center capable of placing live calls to as many as 120 people at once in the victim's own language2

2

. Third, weaponized LLMs trained on malicious data aid cybercrime directly—tools like WormGPT are specifically designed to write phishing emails and malware1

1

. Fourth, AI-augmented malware and infrastructure provide coding assistance and operational support for attacks1

1

. Identity fraud and phishing have benefited most from these developments, with deepfake media and forged documents now indistinguishable from what only skilled operators could produce a year ago2

2

.

Companies Already Losing Millions to AI-Powered Attacks

Companies are already losing millions of dollars to the types of cyberattacks enabled by tools found on these underground markets. "It really goes to show you how cheap it is to conduct these activities...which should be scary for organizations," Kaiser told Axios1

1

. The consequences are concrete: someone with no technical skill, no language fluency, and no operational experience can now buy phishing content, deepfake media, and forged documents that were previously only accessible to skilled operators2

2

. This democratization of hacking capabilities is transforming far more criminals into dangerous hackers1

1

. AI is largely helping criminals perform discrete elements of existing attacks more cheaply and efficiently, though Kaiser noted that "what you're not seeing for sale is an agent that does it all for you and has this amazing success rate"1

1

.

The Ecosystem Preys on Itself While Autonomous Attacks Remain Absent

Halcyon researchers did not find evidence of criminals selling AI systems capable of running fully autonomous AI-driven cyberattacks. Even in AI-enabled malware and infrastructure, much of what researchers observe involves coding assistance, infrastructure, and other individual tasks rather than AI creating and deploying malware independently1

1

. Ironically, the ecosystem also preys on its own participants: infostealers harvest credentials for the very AI tools criminals depend on, and the market's dominant platform was itself breached, exposing 19,000 paying customers2

2

. This recursive dynamic—where tools built to victimize others now victimize the people who buy them—represents a vulnerability that defenders and law enforcement have barely begun to exploit2

2

. The question is no longer whether criminals will adopt AI, but whether defenders can detect, disrupt, and deter an ecosystem that has already reached commercial maturity and scales as fast as the industry it feeds on2

2

. Organizations must watch for continued evolution in social engineering attacks, the potential emergence of more sophisticated autonomous capabilities, and the expansion of freemium models that lower barriers even further for aspiring cybercriminals.