Microsoft Warns of AI Email Attacks Hiding Malicious Commands in Innocent Messages

How 2 Israeli newsrooms covered this story — translated into English and compared side by side.

100% center

Center 2

By דנה גוטרזון

What happened

Microsoft warns of a new cyberattack method where malicious commands are hidden in innocent-looking emails to manipulate AI agents. These attacks, called Prompt Injection, can cause AI to leak data or perform unauthorized actions. Microsoft has developed defenses in Microsoft Defender for Office 365 to detect and block such threats before they reach AI systems.

01Attackers hide malicious commands in seemingly normal emails to manipulate AI agents.
02Prompt Injection targets AI reading emails, not the human recipient.
03Malicious instructions can cause AI to leak data or alter payment details.
04Microsoft Defender for Office 365 now scans emails to block hidden AI-targeted threats.
05Microsoft uses synthetic attack examples to train detection systems.
06Limiting AI permissions and requiring human approval reduces risk of damage.

Summary translated & synthesized from the sources below by baba. Read each original for the full report.