CYBERSECURITY To push this further, Microsoft unveiled M-DASH, a Multi-Model Agentic Security Harness. The tool uses a network of over 100 AI agents powered by different models to scan for vulnerabilities. These agents actively debate each other to determine which risks are the most critical, providing developers with a highly-prioritised, real-world-tested list of fixes before their code is even deployed.
For example, when the tool was first announced in May, it helped Microsoft discover 12 Windows vulnerabilities and some of them were remote code execution related.
AI and security: two sides of the same coin Just as Microsoft is using AI for security in MDASH, organisations also have to make sure they are using security for AI.“ You’ ve got to secure AI and then use AI for security,” asserts Vasu.“ It’ s a connected loop. So, how do we make sure that we’ re securing AI? First, it’ s about understanding the top challenges. Agent fraud is one of the absolute top challenges, followed by data oversharing, protection under novel tests like jail breaks, and making sure identity and access is protected at all times.
60 August 2026