AI Cyberattacks Push Security Toward AI Defense

Anthropic’s Claude Mythos found a vulnerability that had gone undiscovered for 27 years. More recently, signs of attacks using Artex, an AI-based penetration-testing tool, were detected in South Korea’s financial sector. As AI handles tasks from vulnerability discovery to attack execution, security professionals are calling for defenses to shift toward AI as well.

IBM’s “2026 Cost of a Data Breach Report,” published in July, found that AI was used in about 25% of malicious breaches, an increase of 9 percentage points (56%) from the previous year. The report said AI lowers the time and expertise needed to mount attacks, expanding their reach. Data breaches involving AI cost an average of $6 million, about $1 million more than the overall average of $4.99 million. The study covered 602 affected organizations across 17 industries in 16 countries and regions from March 2025 through February 2026.

Organizations that widely adopted AI and automation in security operations had average breach costs $1.93 million lower than other organizations, and took 65 fewer days to detect and contain incidents. The report said AI can rapidly analyze large volumes of security information and support responses.

Security companies are applying AI at multiple stages before and after an attack. Attack surface management, continuous threat exposure management, and AI penetration testing identify exposed assets and exploitable vulnerabilities, while web and API security block automated attacks. EDR and XDR link and analyze unusual activity across multiple systems, and AI security monitoring selects real threats from large numbers of alerts. Zero trust and access controls limit internal movement and data theft after an intrusion.

Security professionals say human oversight is needed for important actions because AI false positives can disrupt normal work. Jun-ho Hong, a professor in Sungshin Women’s University’s Department of Convergence Security Engineering, advised organizations to build layered defenses instead of relying only on one solution’s detection rate. He said this can force attackers to spend more time and resources after breaking in, while limiting internal spread and damage.

Anthropic’s Claude Mythos found a vulnerability that had gone undiscovered for 27 years. More recently, signs of attacks using Artex, an AI-based penetration-testing tool, were detected in South Korea’s financial sector. As AI handles tasks from vulnerability discovery to attack execution, security professionals are calling for defenses to shift toward AI as well.

IBM’s “2026 Cost of a Data Breach Report,” published in July, found that AI was used in about 25% of malicious breaches, an increase of 9 percentage points (56%) from the previous year. The report said AI lowers the time and expertise needed to mount attacks, expanding their reach. Data breaches involving AI cost an average of $6 million, about $1 million more than the overall average of $4.99 million. The study covered 602 affected organizations across 17 industries in 16 countries and regions from March 2025 through February 2026.

Organizations that widely adopted AI and automation in security operations had average breach costs $1.93 million lower than other organizations, and took 65 fewer days to detect and contain incidents. The report said AI can rapidly analyze large volumes of security information and support responses.

Security companies are applying AI at multiple stages before and after an attack. Attack surface management, continuous threat exposure management, and AI penetration testing identify exposed assets and exploitable vulnerabilities, while web and API security block automated attacks. EDR and XDR link and analyze unusual activity across multiple systems, and AI security monitoring selects real threats from large numbers of alerts. Zero trust and access controls limit internal movement and data theft after an intrusion.

Security professionals say human oversight is needed for important actions because AI false positives can disrupt normal work. Jun-ho Hong, a professor in Sungshin Women’s University’s Department of Convergence Security Engineering, advised organizations to build layered defenses instead of relying only on one solution’s detection rate. He said this can force attackers to spend more time and resources after breaking in, while limiting internal spread and damage.