TLDR: Cybersecurity firms are increasingly integrating AI agents into their operations, fostering human-machine collaboration to combat sophisticated threats. While AI offers significant enhancements in detection and response, it also introduces new vulnerabilities and complexities, necessitating robust security controls and a proactive approach to managing autonomous digital entities.
The cybersecurity landscape is undergoing a significant transformation as firms increasingly leverage artificial intelligence (AI) agents to bolster their defenses through enhanced human-machine collaboration. This strategic shift aims to counter the escalating sophistication and speed of cyber threats, which are now frequently augmented by AI themselves.
AI agents, defined as autonomous digital entities capable of thinking, learning, and adapting at speeds far exceeding human capabilities, are becoming a critical component in the arsenal of both defenders and attackers. The World Economic Forum’s “Artificial Intelligence and Cybersecurity: Balancing Risks and Rewards” report highlights that threat actors are increasingly employing AI to amplify the scale and impact of their malicious activities, including supercharging phishing, identity theft, and zero-day exploits. This trend is underscored by Verizon’s 2025 Data Breach Investigations Report, which noted a doubling of AI-assisted malicious emails over the past two years, from approximately 5% to 10%.
While the adoption of AI in cybersecurity is widespread, with a CyberArk survey revealing that 94% of organizations utilize AI for security enhancements, a significant gap exists in implementing adequate controls. The same survey found that only 32% of organizations have established security controls for their AI tools. This disparity raises concerns about the potential for increased attack surfaces and the exploitation of “shadow AI” – unregulated AI tools within enterprises.
Experts emphasize the urgent need for organizations to treat AI agents as highly privileged machines and humans combined, implementing principles such as least privilege, just-in-time access, and zero standing access. The rise of machine identities, which are expected to outnumber human identities by orders of magnitude, further complicates the security landscape, as these autonomous entities request and access sensitive data.
The challenge for cybersecurity professionals is to build AI agents that can detect and counter the sophisticated, AI-simulated network traffic generated by adversarial AI. This creates a meta-level defense requirement, where defenders must develop equally advanced AI capabilities. As Clarence Hinton, Chief Strategy Officer at CyberArk, noted, “It’s a foot race… we have to sprint to actually get shadow AI under control, to actually protect the AI models, to protect the usage of them.”
Also Read:
- AI Agents Accelerate Zero-Day Exploits, Challenging Cybersecurity Defenses
- Corelight Advances Cybersecurity with Generative AI for Enhanced Threat Detection
Despite the inherent risks, the potential for AI agents to enhance detection, response, and automation in cybersecurity is immense. However, this opportunity comes with the shared responsibility of ensuring robust governance and control, particularly when AI interacts with sensitive data. The industry is actively working towards integrating more comprehensive security protocols within AI platforms, recognizing that proactive measures are essential to harness the benefits of AI while mitigating its emerging threats.
