“Analysis of these two models confirms that attackers are actively using malicious LLMs in the threat landscape,” Unit 42 ...