Advancing AI and generative models create new cyberattack methods including AI agents that automate breaches and evade detection.
“Cyberattacks are increasing in frequency and evolving in nature. We are at risk of attack by a variety of adversaries, including nation states, state-sponsored organizations, opportunistic attacks, and organized crime and hackers, through use of increasingly sophisticated methods of attack, including the deployment of AI to find and exploit vulnerabilities, "deep fakes", long-term, persistent attacks (referred to as advanced persistent threats) and the use of the IT supply chain to introduce malware through software updates or compromised suppliers accounts or hardware. In particular, the advance of AI and large language models has given rise to additional vulnerabilities and potential entry points for cyber threats. With generative AI, threat actors may have additional tools to automate breaches or persistent attacks, evade detection, or generate sophisticated phishing emails or other forms of digital impersonation, doing so quickly and without requiring deep technical understanding of potential exploits. Attackers may develop AI agents to fully automate the attack cycle, which could discover new and unexploited applications and dynamically create ways to exploit these weaknesses faster than security tools can adapt to and detect these new attack methods. In addition, increasing use of generative AI models, including new capabilities offered through Model Context Protocol (MCP) Servers, in our internal systems may create new attack methods for adversaries. Because generative”