Blockchain security firm Slowmist warns of AI code poisoning risk.AI code poisoning involves injecting harmful code into AI models' training data.An incident with OpenAI's ChatGPT led to a user losing $2,500 in digital assets after the chatbot recommended a fraudulent Solana API website.Scammers are manipulating AI models to generate fraudulent outputs, posing a threat to users and undermining trust in AI-driven tools.