Your AI Email Assistant Might Be Spying on You: Unit 42 Finds Malicious Extensions

The Hidden Dangers of AI Browser Extensions

AI-powered browser extensions promise to streamline your workflow—generating email drafts, summarizing articles, or even composing social media posts. But according to a recent investigation by Unit 42, some of these so-called productivity boosters are anything but helpful. Disguised as innocent tools, these malicious extensions are designed to steal your data, intercept your prompts, and exfiltrate your passwords. The very assistant you trust to write your emails could be reading your most sensitive information first.

How They Masquerade as Helpers

These high-risk extensions often appear in official browser stores with convincing names and descriptions, thousands of downloads, and even positive reviews—some of which may be fake. Users install them to save time, unaware that behind the curtain, the extension is capturing every keystroke, every text selection, and every API call. Unit 42 researchers found that the malicious code initiates data collection from the moment the extension is activated, without any clear disclosure in the privacy policy.

Data Theft and Password Exfiltration

The most alarming capability is password exfiltration. Some extensions inject JavaScript into web pages to grab credentials from login forms. Others scrape cookies and session tokens, enabling attackers to hijack accounts without ever knowing the password. Unit 42 documented cases where extensions sent stolen data to remote servers located in jurisdictions with lax data protection laws. This exposes users to identity theft, financial fraud, and corporate espionage, especially if the browser is used for work.

Interception of Prompts

For users who rely on AI for composing sensitive emails, the risk is even more insidious. The extension not only reads the final text but also intercepts the prompts and intermediate drafts you send to the AI model. This means your confidential business strategies, personal conversations, and intellectual property are all siphoned off before the AI even processes them. The intercepted data can be sold on dark web markets or used for targeted phishing attacks—after all, the attackers already know exactly what you're working on.

Imagine drafting a pitch for a groundbreaking product or writing a heartfelt email to a loved one—every word is exposed. The extension may even modify the AI response to include malicious links, turning your assistant into a weapon against you.

How to Protect Your Browser

Staying safe does not require abandoning AI tools altogether, but it does demand due diligence and regular maintenance. Here are actionable steps you can take right now.

Check Permissions

Before installing any extension, review the permissions it requests. A simple email helper does not need access to all websites, to your download history, or to modify data on every page. Be especially wary of extensions that ask for “read and change all your data on the websites you visit.” If the permission seems excessive for the advertised function, do not install it.

Stick to Trusted Sources

Even official browser stores are not immune to malicious uploads, but they remain safer than third-party sites. Look for extensions from well-known developers or companies with a verified history. Check the number of downloads, recent update dates, and read the privacy policy—if it is vague or nonexistent, that is a red flag. Unit 42 recommends using curated extension lists provided by enterprise IT departments or security vendors.

Regular Audits

Periodically review the extensions installed in your browser. Remove any that you no longer use or that seem suspicious. Pay attention to changes in browser behavior: unexpected pop-ups, slower load times, or redirected searches can all signal malicious activity. Consider using dedicated browser security tools that monitor extension behavior in real time.

Keep Everything Updated

Outdated browsers and extensions may have vulnerabilities that attackers exploit. Enable automatic updates for your browser and manually check for extension updates regularly. If an extension is no longer maintained, uninstall it—unpatched extensions are a favorite entry point for malware.

Conclusion

The promise of AI productivity comes with a significant security trade-off. Unit 42's investigation reveals that the line between helpful assistant and hidden spy is dangerously thin. By understanding the risks—especially the silent exfiltration of prompts and passwords—and by adopting the protective measures outlined above, you can continue to benefit from AI tools without compromising your digital life. Stay vigilant, audit your extensions, and remember: if a free tool seems too good to be true, your data might be the real price.