What methodology did Halcyon's Ransomware Research Center use to gather data on the underground AI marketplace?
Halcyon studied nearly 4,000 posts across 77 Telegram channels, 20 dark web forums, and five specialized underground markets between the months of October and May.
Q&A ID 5e6ea3af-123e-4dc7-9ea0-2bcc3150abee
What specific costs were associated with various AI-related illicit items found in the underground markets studied by Halcyon?
Sellers offered access to ChatGPT Plus accounts for as little as 10 cents and jailbreak prompts designed to trick models into breaking their safeguards for as little as 32 cents. These low costs make it significantly cheaper for a wider range of individuals to acquire hacking capabilities.
Q&A ID 7685d0a9-a467-4227-bf39-7f0fdef20279
What did Halcyon's Ransomware Research Center find regarding the autonomy of AI systems currently being sold by cybercriminals?
The research did not find evidence of criminals selling AI systems capable of running fully autonomous cyberattacks. Instead, the AI is currently being used to perform discrete elements of existing attacks more cheaply and efficiently, such as providing coding assistance, infrastructure, and individual tasks, rather than creating and deploying malware entirely on its own.
Q&A ID 0c0fb4d1-2960-44fb-89b8-45aedd10e607
How have cybercriminals changed their methods for selling and distributing AI-related cybercrime tools?
Cybercriminals have completely commercialized the sale and creation of these tools, often utilizing software-as-a-service (SaaS) subscription models and online storefronts. Many sellers have moved from traditional dark web forums to easier-to-manage storefronts hosted on Telegram to promote and offer tools they have tested.
Q&A ID 993b5ade-2fe0-4d53-bf15-c79f32d22f97
What specific types of AI tools are currently being sold in underground cybercrime markets, and what are some examples of these tools?
Hackers are selling four primary types of AI tools: 1) Jailbroken and stolen AI services, including prompts to strip safety guardrails from mainstream models for as little as $10; 2) Identity fraud and phishing tools for impersonating real people; 3) Models trained on malicious data to aid cybercrime, such as WormGPT, which is designed to write malware and phishing emails; and 4) AI-augmented malware and infrastructure, including systems capable of calling up to 120 people at once for scam operations.
Q&A ID db576ce6-e2fb-4ee6-b746-1ca0e832fe4c
According to research by Halcyon's Ransomware Research Center, how significantly did advertisements for AI tools on underground forums increase between late 2025 and February 2026?
Advertisements for AI tools on underground forums and platforms surged from fewer than 50 per month in late 2025 to more than 1,400 per month by February 2026. This increase highlights the rapid commercialization of AI tools by cybercriminals.
Q&A ID eb576d4c-0639-436a-a486-245657e40eec