Yes—an email can contain instructions that are difficult for a person to see but that an AI assistant may still process. If an assistant treats those instructions as trustworthy, an attacker may be able to mislead its response, expose information it can access, or prompt an unintended action. Microsoft has documented both email prompt-injection risks and a phishing campaign that used invisible Unicode characters to evade filters, but the campaign report does not establish that attackers used it to hijack AI assistants.
What is prompt injection in an email?
Prompt injection is an attempt to manipulate an AI system by placing malicious instructions in content it processes. When those instructions arrive indirectly in material such as an email, webpage, or document, the technique is called indirect prompt injection. The attacker does not have to message the assistant directly: the assistant may encounter the instructions while doing a task for its user.
As an Amazon Associate I earn from qualifying purchases.
That differs from ordinary phishing, which primarily tries to manipulate a person into clicking, sharing information, or taking another action. An email can target both the recipient and an AI assistant that reads the message on the recipient’s behalf. OpenAI describes prompt injection as a form of social engineering aimed at conversational AI.
How can an email hide instructions from a person?
Invisible Unicode characters
Microsoft Security Research reported a high-volume phishing campaign that used invisible Unicode tag characters in the range U+E0000 to U+E007F. These characters may not appear in typical fonts or interfaces, even though software processing the raw message can receive them. An AI system ingesting raw text may decode characters that a person does not see; the same technique can also disrupt an email filter’s parsing.
#1 Best Overall
- POWERFUL SECURITY KEY: The Security Key C NFC is the essential physical passkey for protecting your digital life from phishing attacks. It ensures only you can access your accounts.
- WORKS WITH 1000+ ACCOUNTS: Compatible with Google, Microsoft, and Apple. A single Security Key C NFC secures 100 of your favorite accounts, including email, password managers, and more.
- FAST & CONVENIENT LOGIN: Plug in your Security Key C NFC via USB-C and tap it, or tap it against your phone (NFC) to authenticate. No batteries, no internet connection, and no extra fees required.
- TRUSTED PASSKEY TECHNOLOGY: Uses the latest passkey standards (FIDO2/WebAuthn & FIDO U2F) but does not support One-Time Passwords. For complex needs, check out the YubiKey 5 Series.
- BUILT TO LAST: Made from tough, waterproof, and crush-resistant materials. Manufactured in Sweden and programmed in the USA with the highest security standards.
Microsoft said the reported campaign used the characters to split financial lure words such as “funding” and make them harder for filters to parse. The researchers described this as an inversion of the familiar prompt-injection use of invisible text: in this campaign, the reported use was filter evasion, not a demonstrated attempt to make an AI reveal mailbox data.
Other concealment techniques
Microsoft’s email-protection documentation also identifies hidden, invisible, or off-screen text; HTML markup and styling; quoted or forwarded thread content; and encoded or obfuscated segments as places where malicious instructions may be concealed. The visible message body is therefore not necessarily the whole message an AI or security system processes.
What could happen if an AI assistant follows the instructions?
The consequences depend on what the assistant can read and do. Microsoft lists possible outcomes including sensitive mailbox content being disclosed, a malicious message being classified as safe, a misleading summary, or an unwanted action in an automated workflow. These are conditional risks—not automatic results of receiving a message.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Rank #2
- POWERFUL SECURITY KEY: The YubiKey 5 NFC is the most versatile physical passkey, protecting your digital life from phishing attacks. It ensures only you can access your accounts
- WORKS WITH 1000+ ACCOUNTS: Compatible with popular accounts like Google, Microsoft, and Apple. A single YubiKey 5 NFC secures 100+ of your favorite accounts, including email, password managers, and more
- FAST & CONVENIENT LOGIN: Plug in your YubiKey 5 NFC via USB and tap it, or tap it against your phone (NFC), to authenticate. No batteries, no internet connection, and no extra fees required
- MOST SECURE PASSKEY: Supports FIDO2/WebAuthn, FIDO U2F, Yubico OTP, OATH-TOTP/HOTP, Smart card (PIV), and OpenPGP. That means it’s versatile, working almost anywhere you need it
- PRIMARY & SPARE KEYS: Just like having a spare house key, we recommend buying two YubiKeys - one for daily use and one as a spare. That way you’ll never get locked out of your accounts
OpenAI has described a security demonstration in which an email encountered during an inbox task redirected an agent toward sending a resignation email instead of completing the user’s requested out-of-office task. This was a demonstration, not a report of a real-world victim incident. It illustrates why an assistant with permission to send messages or use connected tools presents a different level of exposure from one that can only summarize a single email.
The practical risk grows with the assistant’s access: broad mailbox or file access gives malicious instructions more private material to target, while permission to send messages, edit files, or call other tools can turn a misleading instruction into an action. Runtime defenses remain important because the relevant context includes the assistant’s current permissions, tools, and grounded data.
What do the published figures show?
Google Threat Intelligence reported a 32% relative increase in detections in the malicious category when it compared repeated scans of public-web Common Crawl archives from November 2025 with scans from February 2026. Google said the scans did not cover major social media sites and characterized the observed attempts as low in sophistication.
Rank #3
- Security Key : Protect your online accounts against unauthorized access by using FIDO2 and U2F authentication with T110. It's the world's most protective security key that works with windows, Mac OS, Linux as well as Chrome, Firefox, Edge and many other major browsers.
- Certified with the new FIDO2 standard, T110 provides the benefit of fast login and strong protection against phishing, account takeover as well as many other online attactks.
- Works with : Bank of America, Github, Google, Microsoft, DUO, Twitter, Facebook, Dropbox, Apple, ebay, BINANCE, mor and more.
- Fits USB-A port : Insert the T110 security key into the USB-A port of each service and log in conveniently with one touch
- For the driver download and user guide, please visit TrustKey Solutions Home support page.
That figure describes a change in detections within a particular public-web dataset. It is not a count of successful email compromises, an estimate of how many organizations experienced email prompt injection, or evidence that every detected attempt worked. The cited sources do not establish a representative statistic for the number or share of organizations affected by successful email-based prompt-injection compromises.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Which defenses address which part of the risk?
No single control covers the whole chain. Email filtering can inspect suspicious content before delivery; AI-runtime safeguards can constrain how an assistant interprets and uses content; access limits and human approval can reduce the harm if earlier controls miss an attack.
| Defense layer | What it does | What it does not establish |
|---|---|---|
| Email delivery | Microsoft says Defender for Office 365 evaluates inbound messages in its filtering pipeline, including hidden text, HTML, quoted or forwarded content, and normalized encoded material. Its detection uses multiple signals, such as sender reputation, evasion techniques, broader context, and instruction intent. | Microsoft documents a defined detection scope, not universal blocking. Its current focus includes instructions to exfiltrate data through a URL, reveal system prompts or configuration, or discover available tools. It is not intended to block every instruction-like phrase or serve as a general-purpose prompt-injection benchmark. A basic test prompt may not trigger detection. |
| AI runtime | Microsoft describes safeguards such as input filtering, separating user content from system instructions, grounding boundaries, and output filtering. These controls operate where the AI product processes content and makes decisions. | Mail filtering does not replace runtime safeguards. A vendor’s description of its safeguards is not, by itself, an independent evaluation showing that an assistant cannot be manipulated. |
| Permissions and human review | OpenAI advises limiting agents to the access needed for a task, giving specific instructions, and carefully checking proposed actions before confirmation. CIS recommends human approval before AI executes code or makes high-impact changes, limiting access to sensitive systems and data, keeping inventories of accessible data, systems, and tools, and training staff about prompt-injection risk. | These measures constrain access or impact; they do not prove that an email or instruction was detected. |
What should users and organizations do?
Limit what the assistant can access
Grant an email assistant only the mailbox, files, and tools required for its task. Prefer a narrow request such as summarizing a specified message over a broad instruction to review email and take whatever action is needed. The narrower the task and access, the fewer opportunities a malicious message has to influence unrelated work.
Rank #4
- POWERFUL SECURITY KEY: The YubiKey 5C NFC is the most versatile physical passkey, protecting your digital life from phishing attacks. It ensures only you can access your accounts
- WORKS WITH 1000+ ACCOUNTS: Compatible with popular accounts like Google, Microsoft, and Apple. A single YubiKey 5C NFC secures 100+ of your favorite accounts, including email, password managers, and more
- FAST & CONVENIENT LOGIN: Plug in your YubiKey 5C NFC via USB and tap it, or tap it against your phone (NFC), to authenticate. No batteries, no internet connection, and no extra fees required
- MOST SECURE PASSKEY: Supports FIDO2/WebAuthn, FIDO U2F, Yubico OTP, OATH-TOTP/HOTP, Smart card (PIV), and OpenPGP. That means it’s versatile, working almost anywhere you need it
- PRIMARY & SPARE KEYS: Just like having a spare house key, we recommend buying two YubiKeys - one for daily use and one as a spare. That way you’ll never get locked out of your accounts
Require approval for consequential actions
Have a person review proposed messages, file changes, code execution, and other high-impact actions before the assistant carries them out. Treat an instruction found inside an email as untrusted content, not as authorization to override the user’s request or the system’s rules.
Use layered safeguards
For organizations deploying email assistants, combine mail-channel inspection with safeguards in the AI workflow: keep user content separate from system instructions, constrain what data can ground a response, filter outputs, and review actions. Microsoft Defender for Office 365 is one documented enterprise email-security example, but its stated scope should be checked against current product, licensing, regional, and availability details before adoption; it should not be treated as a guarantee against all prompt injection.
Recommended Free Tools
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
