BRK325 Breakout Wed 18 Nov 11:30 PST
Behind the AI models built for security at scale
Anthony Penta, Nazmus Sakib
- Microsoft session page (opens in a new tab)
- Watch on-demand (opens in a new tab)
- Slides: rights-protected, available via the Microsoft session page (opens in a new tab)
Summary
From the recording / Microsoft AI summary
The session covered Microsoft's MAI Cyber models, built for security tasks with cost per task as the headline goal. Speakers said they are not trained on customer data. Sign-up for the M-dash preview is available, but it is not broadly available. Pricing and GA dates were not stated.
Session description
Microsoft's catalogue description
Always-on defense demands AI that runs continuously, but accuracy, latency, cost, and trust can keep powerful models out of security workflows. See how purpose-built models like MAI-Cyber-1-Flash pair deep cyber specialization with rigorous evaluation to deliver strong security reasoning efficiently, optimize AI economics, and maintain safety and governance from training through deployment. Learn how specialized models strengthen vulnerability discovery and analysis at scale.