> Detected for the first time, malware attempts AI evasion by injecting a prompt to tell the LLM to label the file as benign