DSIT: Thematic review and gap analysis on AI security

The UK Department for Science, Innovation and Technology has published a comprehensive gap analysis of AI security research covering 9,000+ peer-reviewed publications (2021–2026), identifying 12 core security themes and mapping critical blind spots including training-data assurance, model provenance, agentic-AI systems, and safe model decommissioning. The UK government has just mapped the exact security gaps in AI systems that vendors are unprepared for: training-data assurance, model provenance, and agentic-AI safety are all flagged as critical blindspots with no mature solutions. If you're building tools to verify training data, track third-party model lineage, or control autonomous agents, you have a regulatory tailwind and proven customer need.