Proofpoint: Generative AI Erases Behavioral Tells in Malicious Emails

Proofpoint stated at its Protect 2026 conference that generative artificial intelligence is fundamentally altering the appearance of malicious email campaigns. The company argued that the behavioral signals traditional email filters have relied upon for years are fading because attackers can now produce fluent, unique messages for every recipient.

The vendor highlighted a specific campaign where a single threat actor contacted more than 30 real estate agents at one organization. Each agent received a distinct message regarding a different property, written in different words, yet every thread concluded with the same request to join a video meeting.

The Erosion of Detection Signals

Traditional email security tools were designed for attackers who had to cut corners due to time constraints. Researching targets, writing convincingly in a second language, and producing enough variants to defeat signatures required significant effort. Consequently, attackers reused templates and infrastructure, leaving behind detectable patterns such as new infrastructure, awkward language, or odd timing.

Generative AI removes these constraints. According to Proofpoint, attackers can now generate language that is fluent and localized in any language without typos or translation artifacts. They can create context-aware replies that fit the recipient’s role and the specific timing of the conversation.

Furthermore, threat actors can match writing styles and signature blocks while fabricating entire conversation histories. A unique message can be generated for every recipient in seconds, eliminating the template reuse that behavioral detection systems previously used to identify threats.

Federal Data and Industry Impact

Federal data supports the trend of increasing sophistication and volume in these attacks. The FBI’s Internet Crime Complaint Center counted more than 22,000 complaints involving AI in 2025. These incidents resulted in nearly $900 million in associated losses, according to the FBI’s 2025 Internet Crime Report.

The same report logged 24,768 business email compromise complaints with approximately $3.05 billion in reported losses. This data underscores the financial impact of campaigns that bypass traditional detection methods by appearing normal and personalized to each target.

Proofpoint argues that because AI allows attackers to make every message look normal, the analysis methodology must change. Security teams can no longer rely solely on asking whether a message looks unusual. Instead, they must focus on determining what the message is trying to achieve.

What should you do?

Security professionals should review their current detection rules to ensure they are not relying exclusively on linguistic anomalies or template matching. Organizations should evaluate their email security posture to see if it can identify requests for unusual actions, such as video meetings, regardless of the message's fluency.

Teams should monitor for high volumes of unique messages targeting specific departments or roles, as this may indicate an AI-driven campaign. Users should be trained to verify requests for sensitive actions through secondary channels, even if the email appears personalized and contextually accurate.

Background: Generative AI

Generative AI examples include text completion, image synthesis and code generation. These systems predict the next likely token based on vast training datasets. Understanding the underlying mechanics helps you identify risks like data leakage and hallucination in your infrastructure.

Read the full guide: Generative AI Examples: Mechanisms, Risks and Control

What to do and how to stay safe: Proofpoint

• Review email filtering rules to reduce reliance on linguistic errors or template matching as primary indicators of compromise.
• Implement verification procedures for requests involving financial transactions or external meetings, regardless of sender reputation.
• Monitor for spikes in unique, personalized messages targeting specific employee roles or departments.
• Train employees to recognize that fluent, well-written emails can still be malicious and require verification.

Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

How is generative AI changing email security?

Generative AI allows attackers to create fluent, personalized, and unique messages for each recipient, removing the behavioral tells like clumsy wording that filters previously used to detect threats.

What is the financial impact of these attacks?

The FBI reported nearly $900 million in losses from AI-related complaints and about $3.05 billion in business email compromise losses in 2025.

What example did Proofpoint provide?

Proofpoint described a campaign where one actor sent more than 30 distinct, tailored property inquiries to real estate agents, all ending with a request to join a video meeting.