AI speeds up cyberattacks and puts defenders under pressure
AI is transforming cybersecurity by allowing attackers and defenders to inspect vast quantities of software code rapidly. Agentic systems can automate vulnerability searches that previously required extensive manual work. Google has reported attackers using these systems to examine software updates, identify the security problems they address and develop ways to exploit them.
The technology primarily accelerates discovery of familiar weaknesses rather than entirely new classes of flaws. Researchers have used AI to uncover vulnerabilities in Java and PHP software, although University College London security researcher Steven Murdoch cautions that the Java finding’s significance has been exaggerated. Human failures, including weak passwords and clicks on malicious links, remain central to most breaches. AI can also make phishing more convincing by personalizing messages with publicly available information.
Defenders are using AI to inspect more software, and researchers are developing agents that can detect and repair vulnerabilities automatically. But faster exploitation leaves organizations less time to respond after flaws become public. Rushed patches can interfere with other software and cause outages themselves, complicating the response.
Recent disruptions affecting Microsoft, US air traffic control, JPMorgan Chase and the US National Security Agency have heightened infrastructure concerns. Murdoch considers a coordinated cyberattack unlikely and says the most prominent incidents were probably technical malfunctions.