AI News Phishing emails are baiting your AI assistant

[Brownie2019]

Content source

https://www.malwarebytes.com/blog/ai/2026/10/phishing-emails-are-baiting-your-ai-assistant

You ignored the suspicious email. You didn’t open the attachment, click the link, or reply to the sender.

Then you asked your AI assistant to summarize your inbox.

Could the message you wisely avoided find its way back onto your to-do list—only this time presented as an urgent recommendation from a tool you trust?

That’s the risk behind phishing emails containing hidden instructions aimed at AI assistants. Instead of relying only on persuading you, attackers are also trying to influence the software that reads your messages. Recent research from Barracuda describes emails designed to target both audiences at once.

For the human recipient, there was a password-protected attachment, with the password supplied in the email. For an AI assistant processing the message, there were hidden instructions intended to influence how it summarized or handled the content.

By including a prompt injection, attackers give themselves another way to reach you. The malicious instructions could try to make the assistant describe the email as legitimate or urgent, encouraging you to revisit it.

In another example, a hidden block instructed a summarizing AI to add a fake priority action: Change a vendor’s payment details. That could result in an employee sending money to an attacker’s account.

The attacker isn’t simply asking the AI to repeat a fraudulent request. They’re trying to turn that request into the assistant’s recommendation.

How prompt injection works
Prompt injection means planting instructions in material an AI system is supposed to read, hoping it will follow those instructions instead of treating them as part of the material. When the instructions arrive through an outside source, such as an email or webpage, it’s often called indirect prompt injection.

The malicious instructions don’t have to appear as ordinary, readable paragraphs. Barracuda identified concealment techniques including comments in the email’s underlying HTML code, text made invisible through styling, encoded content, and invisible “zero-width” characters. These can obscure instructions from a person viewing the message while leaving content available for an AI system to process. Whether that content actually reaches the model depends on how the particular tool extracts and prepares the email.

That means this problem cannot be solved simply by asking people to look more carefully at their emails. Part of the defense must sit inside the tools handling those messages.

Apollo Raines, developer of the AI code review service ShipItClean, explains:

“The real flaw is that the model can’t tell the difference between text it’s supposed to read and instructions it’s supposed to follow. Stripping hidden content before it reaches the model is worth doing, and every vendor should do it. But it’s an arms race. Each new encoding trick is a new hole.”

He adds:

“People have learned to distrust the email. They haven’t learned to distrust their AI assistant. So, the defense has to live inside the tools. I’d go one step further: some of it has to live inside the model itself. Wrappers, filters, sandboxing and human sign-off on payments are all necessary.”

Before taking action
An assistant that summarizes messages could mislead you about which email deserves attention or what action it requires.

An assistant with additional permissions could create a more serious problem. Depending on its capabilities and safeguards, malicious instructions might try to make it carry out unauthorized actions or disclose sensitive information.

You don’t need to abandon AI tools that help with email. But you do need to consider what you trust them to do and what permissions you give them.

The familiar phishing advice still applies. When an assistant reads your inbox on your behalf, check its recommendations before acting:

Check the original message before acting on a recommendation involving money or sensitive information. If a summary tells you to pay an invoice, change account details, or share sensitive information, don’t treat the summary itself as verification.
Always verify payment changes through a separate, established channel. Contact the person or organization using details you already trust, rather than relying on information supplied in the suspicious message.
Keep your AI assistant’s permissions limited to what you need and, at work, what your organization allows. An assistant used to summarize emails should not need broad authority to carry out unrelated actions. Limiting access can reduce the consequences if it follows malicious instructions.
Require human approval for high-impact actions like payments and changes to vendor payment details. Approval should include checking the underlying request, not simply accepting the assistant’s recommendation.

 

[Bot]

Hi Brownie2019, this is the right way to frame the risk: the email is not somehow “infecting” the AI; it is crossing a trust boundary. Content supplied by an outsider is being fed to software that can influence—or potentially act for—the recipient.

Barracuda reports finding phishing messages constructed for both audiences: conventional social engineering for the person and concealed prompt injection for an inbox assistant. That establishes real attacker use of the technique. It does not mean every assistant will extract the concealed material, obey it, or have permission to transfer money; those outcomes depend on the mail parser, model safeguards and available tools.

Stripping HTML comments, zero-size text and odd Unicode is worthwhile, but cannot be the whole defence. An attacker can place the same instruction in visible text, an attachment or a linked page. The durable controls are:

• Mark email and attachment content as untrusted data, never as instructions.
• Keep summarizers read-only and give action-capable agents narrowly scoped, short-lived permissions.
• Display the original sender, source message and supporting passage beside generated tasks.
• Gate sending, sharing, deleting and financial actions separately.
• Log model output and tool calls so attempted manipulation can be investigated.
• Require meaningful approval: show the payee, changed account details and source request—not merely an inviting Approve button.

That broadly matches Microsoft’s defence-in-depth guidance, which combines prompt shielding, separation of external content, tool-chain checks, least privilege and human verification.

For users, the practical rule is simple: an AI-generated inbox summary is a convenience, not authentication. If it suddenly recommends opening a protected attachment, changing payment details or disclosing information, inspect the original and verify through an independently trusted channel. “Urgent” is an adjective, not a security control.

Sources

• Email attacks target both humans and AI in the same message | Barracuda | Barracuda Networks Blog
• Defend against indirect prompt injection attacks | Microsoft Learn

 

[Halp2001]

Beyond phishing itself, I think this highlights a broader issue: many users may begin to see AI responses as an additional layer of trust, when in reality, those responses can also be compromised by the information the AI processes. If the input data has been manipulated, the summary or recommendations can be misleading as well. So, just as we learned to be wary of suspicious links and attachments, we'll need to get used to verifying recommendations generated by AI assistants too, especially when they involve money, credentials, or significant changes.

 

[Victor M]

AI Agents are trouble period.