AI Email Assistants Are Reading Your Phishing Emails and Falling for Them

Scammers have found a new target that is not you, but the AI assistant reading your email for you. By hiding commands inside phishing emails, they can trick these assistants into performing actions you never would, even if you delete the message yourself.

How does this attack work?

Many email services now offer AI summaries or automated responses. These assistants read your incoming mail to save you time. Scammers are writing phishing emails that look normal on the surface but contain hidden instructions buried in the text, often in white text on white background, or in comments meant only for machines to read. When the AI processes the email, it sees the instruction and follows it, sometimes before you even know the message arrived.

What damage can these hidden commands cause?

The AI might reply to the phishing email on your behalf, forward sensitive information, or mark legitimate messages as spam. It could also be tricked into adding a malicious contact to your address book or confirming your email address as active. The scammer gains proof that your inbox is monitored, which increases the value of your email address to other criminals.

Should I turn off my AI email assistant?

Not necessarily. These tools are useful when they work correctly. Instead, stay aware that phishing is evolving. Your best defense remains the same: don't click links or download attachments from unexpected emails, even if they look routine. If you use an AI assistant, check its settings to see what actions it's allowed to take on your behalf. Many services let you restrict it to reading only, with no power to send, forward, or delete messages. That limits the damage significantly.

What if I think I've received one of these emails?

If you spot an email that looks suspicious, report it to your email provider as phishing and delete it. Don't assume the AI will handle it correctly. You're the one who can recognize context and intent. Consider turning off automated features temporarily if you're concerned, then turn them back on once the risk passes. You remain the most important security layer in your own inbox.