AI expands cyberattack capabilities and defensive tools

AI is reshaping cybersecurity by giving attackers and defenders new ways to automate tasks, analyze large amounts of data and discover vulnerabilities. Agents can accelerate offensive operations, while defensive applications include threat detection, incident analysis and response.

Security research spans alignment failures in the UK AISI cybertesting incident, the discovery and exploitation of vulnerabilities in critical software with Claude, and different levels of AI use in attacks targeting Japan. The Australian Signals Directorate has warned that disruptions are inevitable as AI-powered cyber threats grow. Other areas of scrutiny include the use of AI tools for doxing and unconventional surveillance through invasive advertising.

Risks also extend to the software ecosystem surrounding AI. Examples include attacks using signed ChatGPT and Claude binaries and the return of the Load/Mine malware campaign on Hugging Face after a takedown. Sandboxed development environments and assessments of models’ potential to self-replicate are further areas of security investigation.