Hidden Prompts Trick AI Into False Email Summaries

With some simple HTML that's invisible to users, attackers can manipulate AI-powered email summarizers into producing malicious information.