Security Affairs rounds up AI security news and research

Security Affairs published the fifth edition of its weekly AI-cybersecurity newsletter on 25 October 2026, collecting links to coverage and research on artificial intelligence and security. It describes AI as changing both sides of cybersecurity: agents can automate tasks, analyse data, find vulnerabilities and speed up offensive activity, while AI tools can support threat detection, incident analysis and response.

The collection includes reports on alleged AI-assisted attacks and security risks, including a supply-chain compromise attributed by Microsoft to the Storm-0940 group using its ARTEX agentic AI, and a credential stealer and backdoor delivered through a compromised LiteLLM package. Other links cover AI safety and monitoring, the security of frontier-model training environments, deepfake detection for UK elections, and proposals for UK AI security funding. The newsletter provides headlines and links rather than details on methods, evidence, affected systems or recommended responses.