LLMs are Vulnerable to Malicious Prompts Disguised as Scientific Language