IN SecurityWeek’s interview with IBM UK’s Steve Wallin, the impact of AI on cyberattacks is framed as a speed and scale problem rather than a new technique. Phishing remains the leading attack vector, accounting for 17% of breaches and averaging $5.29 per breach according to the IBM Cost of a Data Breach Report 2026. AI enhances phishing through more personalised content, while also helping attackers identify vulnerabilities, exploit the phished entry point, and elevate privileges for rapid lateral movement.
Although AI-driven attacks are still a minority today (about 20% of incidents), Wallin warns that the likelihood and speed of breaches will rise unless defenses are redesigned.
The article stresses that the fundamentals of cybersecurity have not changed with AI; what shifts is the tempo. Defence must become proactive, with “security by design” and a strong emphasis on zero trust to control credentials, access and data. John Kindervag, the “father of zero trust,” is cited confirming that properly implemented zero trust can still stop AI-enabled packets as they traverse networks.
Beyond prevention, resilience and recovery are crucial, including rapid data restoration through cyber vaults and immutable backups. The piece cites IBM’s Z mainframe cyber vault as an example of automated validation and forensic tools for rapid recovery. In short, AI raises the urgency of prioritised risk management, automation, and continuous improvement of security postures, rather than introducing new defence paradigms.