The fast advancement of AI technology presents an emerging and serious challenge: AI compromise. Cybercriminals are steadily exploring methods to exploit AI platforms for malicious purposes. This encompasses everything from poisoning learning data to evading security protections and even deploying AI-powered attacks themselves. The potential effect