AI Attackers Will Soon Act Like Insiders

TL;DR: Proofpoint's CEO warns that new AI-powered cyberattacks will act like trusted insiders, bypassing traditional defenses. This requires a fundamental shift in how companies approach security against automated, intelligent threats.
Key facts
- Category
- AI
- Impact
- High
- Published
- Source
- TechRadar
Full summary
Proofpoint's CEO warns that new AI-powered cyberattacks will act like trusted insiders, requiring a fundamental shift in security strategy.
Proofpoint CEO Sumit Dhawan has issued a stark warning about the next evolution of cyberattacks, urging companies to prepare for a new class of threats driven by agentic AI. Speaking at a recent company event in London, Dhawan advised security leaders to begin thinking of these AI systems as potential “insiders in your environment.” According to reporting from TechRadar, this perspective marks a critical departure from traditional threat models that focus primarily on external attackers. The warning highlights a near-future scenario where autonomous AI agents, not just human hackers, are the primary actors in sophisticated security breaches. This shift implies that the speed, scale, and nature of attacks are about to change dramatically, forcing a complete re-evaluation of defensive strategies across every industry.
The term “agentic AI” describes autonomous systems that can pursue high-level goals with minimal human intervention. In a cybersecurity context, this is a significant leap beyond current AI-assisted attacks, which might use language models to craft more convincing phishing emails. An AI agent, by contrast, could be given a simple objective like “exfiltrate customer financial data.” From there, it could independently probe a company’s network, identify and exploit vulnerabilities, move laterally between systems, escalate its own privileges, and adapt its methods in real-time to avoid detection. Because these agents can learn and reason, their behavior would be far less predictable than that of scripted malware. They could mimic legitimate user activity over long periods, making them exceptionally difficult to spot with conventional security tools that look for known patterns of attack.
This development fits into a broader trend of an accelerating AI arms race in cybersecurity. While attackers are developing offensive AI tools, security firms are simultaneously building defensive AI to hunt for threats and automate responses. The move toward agentic systems on both sides means the speed of conflict is increasing exponentially, shrinking the window for human intervention from days or hours to mere seconds. Dhawan’s “insider threat” analogy is particularly potent because it underscores the limitations of perimeter-based security. Firewalls and intrusion detection systems are designed to keep external threats out, but they are far less effective against an attacker that already has a foothold inside the network. An AI agent that successfully compromises a single low-level account could quickly become a powerful and persistent internal threat.
For CTOs, developers, and security teams, this warning serves as a call to action to modernize their defensive postures. Relying on signature-based antivirus software and perimeter firewalls alone is no longer a viable strategy. The focus must pivot to a zero-trust architecture, where no user or device is trusted by default, and to advanced behavioral analytics. Security systems need to be capable of identifying anomalous activity from any account, whether it belongs to a human or an AI. This means investing in robust Identity and Access Management (IAM), Privileged Access Management (PAM), and modern Security Information and Event Management (SIEM) platforms. These tools can correlate subtle signals from across the network to build a picture of user behavior and flag deviations that might indicate a rogue AI agent at work. The future of defense lies not in blocking known threats, but in understanding intent and spotting the unfamiliar.
Why it matters
The shift from AI-assisted attacks to autonomous AI agents represents a fundamental change in threat models. For engineers and security architects, this means legacy perimeter defenses are becoming obsolete. Defending against these threats requires a move towards zero-trust principles and advanced behavioral analytics inside the network.
Business impact
Agentic AI attacks can operate at a scale and speed that human-led security teams cannot match, increasing the risk of rapid, widespread data breaches. Businesses must invest in next-generation security tools and strategies to counter these automated threats or face significant financial and reputational damage.
Tags
Related on Notifire
Related stories
Primary source: TechRadar