Palo Alto Networks
By Michael Sikorski, CTO and VP of Engineering, Unit 42, Palo Alto Networks
The cyber threat landscape is reshaping under the influence of AI. At Palo Alto Networks we look at AI’s impact on cybersecurity in two main ways: An acceleration of attacks, with adversaries leveraging AI to operate with speed and sophistication; and an expansion of the attack surface, with AI introducing new attack vectors for adversaries to target and leverage to infiltrate global organizations.
The Rise of the AI-Powered Attacker
Attackers are weaponizing AI to augment every stage of their operations. Our 2025 Unit 42 Global Incident Response Report reveals that attackers can exfiltrate data in under five hours in 25% of incidents – in some cases in under an hour. Unit 42 researchers demonstrated that AI-assisted attacks could reduce the time to exfiltration to under 25 minutes.
AI-driven methods are accelerating the attack lifecycle, making threats harder to detect and faster to execute. Attackers are using LLMs to fuel convincing and scalable phishing campaigns, conduct reconnaissance, and create deepfakes that are increasingly employed in advanced social engineering attacks. This is not just about faster attacks, but a fundamental shift in the threat paradigm.
The Agentic AI Vulnerability Frontier
Agentic AI, with its autonomous decision-making and execution, introduces new vulnerabilities and further expands the attack surface. We are confronting risks like prompt injection, where malicious instructions can override an agent's intended purpose and lead to unauthorized actions or data breaches. Unauthorized autonomous actions become a significant concern as agents, if misled or hijacked, could perform unintended operations like modifying access or approving fraudulent transactions. Data exfiltration and leakage are also pressing issues, as agents processing sensitive data might be tricked into revealing confidential information through carefully crafted queries. As Agentic AI revolutionizes business processes, our security approach must fundamentally shift, securing this new frontier while defending against increasingly quick adversaries.
Building Resilience in the AI Era
Palo Alto Networks understands that securing this evolving AI ecosystem demands more than traditional defenses. This involves extending established security principles, while building essential, new AI-specific guardrails. This requires integrating security from the initial stages of AI agent development, applying strict access controls and segmentation, rigorously validating inputs, and continuously monitoring agent behavior.
Our comprehensive approach directly addresses these challenges:
We gain insights and expertise from over 500 Unit 42 incident response investigations annually and over 8.9 million new and unique attacks blocked daily, allowing us to continually update our playbooks against emerging threats. The time to act is now to ensure AI benefits, rather than undermines, our shared future.
Visit Palo Alto Networks at Booth #3240 at Black Hat USA to learn more.