Evaluation of LLM Vulnerabilities to Being Misused for Personalized Disinformation Generation