The UK and Ukraine have established a partnership granting British researchers and companies access to Avengers AI Labs, a platform containing extensive battlefield data collected through thousands of cameras and sensors that trains AI models to recognize military targets and currently analyzes over 100,000 drone video feeds monthly. The collaboration aims to combine Ukrainian operational insights and real-world datasets with British technological expertise, with pilot projects already underway involving British startups developing AI-enabled sensor systems for base protection and next-generation low-power AI chips for autonomous systems. The partnership, formalized between PM Andy Burnham and President Zelenskyy, is positioned as "AI sovereignty in practice" and represents part of the broader UK-Ukraine 100 Year Partnership focused on defence and national security innovation.
Read full article at The Cyber Express ↗The Australian Signals Directorate has outlined its 2026 board priorities alongside new frontier AI guidance, emphasizing that rapid response alone is insufficient to counter emerging cyber threats in an AI-driven threat landscape. The directorate's approach suggests a need for more comprehensive strategies beyond speed-focused defenses to address the evolving sophistication of AI-enabled attacks.
The Linux Foundation has adopted TRACE, a hardware-backed specification developed collaboratively by AMD, Intel, Microsoft, OPAQUE, and the Technology Innovation Institute to create standardized evidence records for AI agents and confidential workloads. TRACE provides organizations with a consistent, trustworthy method to demonstrate that sensitive data handling complies with established policies as autonomous AI agents and open-weight models become increasingly prevalent in enterprise deployments.
Tricentis' CISO Erika Dean discusses the persistent practice of using production data in test environments and argues that safer alternatives are now sufficiently mature to eliminate the need. She highlights her team's discovery of a prompt injection vulnerability during red-teaming that warranted delaying a release by a week, and outlines criteria for evaluating AI vendors, particularly scrutinizing their vague explanations regarding data location and retention practices.
CISA red-team exercises revealed divergent security capabilities across critical infrastructure sectors, with water utilities successfully detecting and halting simulated attacks while a government organization failed to stop the same intrusion attempts after initial compromise. The findings highlight significant gaps in incident response and threat detection maturity within some government agencies compared to their water sector counterparts. These controlled tests underscore the varying levels of preparedness for advanced persistent threats across critical infrastructure.