Myth #1: AI can only defend against known threats

Modern AI-driven cybersecurity doesn't rely solely on signatures or known attack patterns. Instead, it learns what normal behaviour looks like across networks, systems, and users, and flags deviations that may indicate a threat.

This is why AI can help identify novel attacks, including some zero-day exploits, before traditional signatures are available. During the 2023 MOVEit Transfer attacks, security researchers highlighted the importance of behavioural and anomaly-based detection, as suspicious activity could be observed before signatures were developed.

That doesn't mean AI is perfect. Sophisticated attackers continue to develop techniques designed to evade detection. Recent findings from the 𝘌𝘕𝘐𝘚𝘈 𝘛𝘩𝘳𝘦𝘢𝘵 𝘓𝘢𝘯𝘥𝘴𝘤𝘢𝘱𝘦 (2025) also show how adversaries are increasingly leveraging AI tools, reinforcing the need for adaptive and continuously evolving defence mechanisms.

The idea that AI can only recognise threats it has already seen is outdated.

Projects like AIAGENT4CYBER are exploring how multi-agent AI systems can combine anomaly detection, threat intelligence, and contextual reasoning to strengthen cyber resilience in increasingly complex environments.