Google Gemini Flaw Lets Hackers Hijack Email Summaries Without Links or Attachments

If your company uses Google Workspace, it’s time to look closely at how Gemini handles email summaries. Security researchers have uncovered a flaw in Gemini’s summarisation feature that attackers can use to hide phishing commands inside email content, with no links or attachments required.

As CyberSecurityNews reports, the attack works by embedding HTML and CSS into emails. Gemini processes that code when generating summaries, not the actual visible content. Hackers can disguise prompts using invisible styles like white-on-white text or zero font sizes to slip in fake warnings that appear to come from Google itself.

When someone hits “Summarise this email” in Gmail, Gemini can surface a malicious message, for example, urging the user to call a fake support number or visit a phishing site. Since the email itself looks empty or harmless, users are less likely to suspect anything unusual.

Unlike typical scams, this attack doesn’t rely on links or scripts. It’s powered purely by how Gemini interprets hidden markup. Researchers warn this approach could extend beyond Gmail to Docs, Slides, and Drive, raising concerns about automated phishing across the Workspace ecosystem.

Google responds to concerns

According to BleepingComputer, Google has acknowledged the issue and pointed to its broader work to defend against prompt injection attacks.

“We are constantly hardening our already robust defenses through red-teaming exercises that train our models to defend against these types of adversarial attacks,” a Google spokesperson told the news outlet. Not to forget, the company also referenced its blog post outlining how it’s improving protections.