As AI agents are becoming more prevalent, AI agents are becoming more susceptible to phishing and social engineering tactics. Fortifying communications against potential threats Proofpoint, the global leader in digital communications security and compliance, recently launched new AI-powered capabilities at this year’s Proofpoint Protect 2025 virtual summit. This change comes as good news as the company currently scans 3.5 billion emails per day. It examines close to 50 billion URLs and 3 billion attachments, over one-third of the worldwide email traffic.
Proofpoint regularly releases new AI-based capabilities across its market-leading Prime Threat Protection. This upgrade will greatly enhance our ability to detect threats against human users and AI systems. As these AI tools become more integrated into workplace environments, they expand the attack surface, creating new challenges for cybersecurity measures.
Advanced Threat Detection Capabilities
Proofpoint is continuously iterating its threat detection models down to 300 million parameters and retraining every 2.5 days. This constant iteration improves its capability to better understand the intent behind messages. The company uses hundreds of signals, behavioral, reputational and content-based, to flag potential threats.
Real-world insights from Daniel Rapp, Chief AI & Data Officer at cybersecurity company Proofpoint, reinforced the need for this multi-faceted approach. He mentioned that by preventing attacks before delivery, Proofpoint prevents the user compromise and exploitation of the AI systems from ever happening.
“We have placed detection capabilities directly in the delivery path, which means latency and efficiency are critical,” – Daniel Rapp.
Proofpoint’s RFC-822 compliant engines inspect email headers, plain text and HTML content. This enables them to spot signs of fraud that show when a message is a phishing attack. Rapp pointed to a concerning pattern in the recent spate of attacks. The difference between the HTML and plain text versions of messages can be quite extreme, greatly complicating any detection efforts.
The Growing Threat Landscape
As AI agents revolutionize the way enterprises do business, they give cybercriminals a new vector to exploit. Todd Thiemann, cybersecurity analyst at Omdia, makes a key point. He shares that generative AI assistants and copilots significantly multiply the attack surface in enterprises. Conventional security architectures aren’t built to address these new, profound vulnerabilities.
Thiemann stated, “Security tooling must evolve from detecting known bad indicators to interpreting intent for humans, machines, and AI agents.” This change is enormously important. Attackers are quickly finding new ways to exploit AI technologies, employing techniques such as prompt injections—text-based payloads that drive malicious machine reasoning.
“Prompt injections and other AI-targeted exploits represent a new class of attacks that use text-based payloads that manipulate machine reasoning rather than human behavior,” – Todd Thiemann.
The real challenge is not just in identifying these threats but bending our current security safeguards to meet them. The rapid adoption of AI across domestic infrastructure and institutions means that public and private organizations need to reconsider their cybersecurity approaches.
Implications for Organizations
As AI plays an ever-increasing role in an organization’s business operations, companies have to work smarter to protect themselves against evolving threats. Rapp believes Proofpoint’s sophisticated detection technology will help protect humans—and AI counterparts—alike. Through the use of a multi-layered defense that uses several different detection methods, Proofpoint technology helps stop phishing threats before they can do harm.
“This is important because since AI agents are so literal, they are susceptible to attacks,” Rapp continued. These unique properties of AI systems may be subject to novel forms of user manipulation, distinct from typical user interactions.