Warning: Discounted 'Claude' AI Access Sold on Cybercrime Forums Puts User Data at Risk

Cybersecurity researchers have uncovered more than half a dozen underground services advertising illicit access to popular AI models on cybercrime forums and messaging platforms. One such service, dubbed 'Poison Claude', claims to provide discounted access to several versions of Anthropic's Claude models, including Opus 4.8, Opus 4.7, Opus 4.6, and Sonnet 4.6.

What makes this scheme particularly dangerous is that the operator running the service can view every prompt entered by customers using the discounted access. This means any business or individual using such a service unknowingly exposes sensitive information—company data, client details, internal strategies, or proprietary code—directly to the criminal operator behind the scam.

This discovery highlights a growing trend of threat actors exploiting demand for cheap AI access to harvest data from unsuspecting users. For small businesses experimenting with AI tools to boost productivity, using unauthorised or 'too good to be true' discounted AI subscriptions can create a serious data leakage risk, even if the underlying AI brand appears legitimate.