API
< Back to all clusters
[TECHNOLOGY] · United States · 131 sources (showing 100)

started · updated

Microsoft launches MAI‑Cyber‑1‑Flash and Project Perception

Microsoft announced its first cybersecurity‑specific foundation model, MAI‑Cyber‑1‑Flash, and an agentic security platform called Project Perception. The model, built on Microsoft’s MAI‑Thinking‑1 line and integrated into the MDASH harness, scored roughly 96 % on the CyberGym benchmark, surpassing Anthropic’s Mythos, Google’s Gemini and OpenAI’s GPT models. Microsoft says the combination delivers about a 50 % cost reduction versus leading AI security solutions.

Project Perception coordinates red‑team, blue‑team and green‑team AI agents to simulate attacks, detect vulnerabilities, investigate incidents and automate remediation. The platform automates up to 90 % of routine security tasks, routing the most complex queries to larger models such as GPT‑5.4. Both technologies were unveiled at a San Francisco event in late July 2026, with Mustafa Suleyman (co‑founder of DeepMind, head of Microsoft AI) and Hayete Gallot (Microsoft VP for security) presenting the announcements. Project Perception will enter public preview in early August for vetted Azure customers, aiming to defend enterprises against AI‑enabled cyber threats at speed and scale.

Entities

Hayete Gallot · MAI-Cyber-1-Flash · MAI‑Cyber‑1‑Flash · MDASH · Microsoft · Microsoft Corporation · Mustafa Suleyman · Project Perception

Claims

What the coverage asserts, and how many sources carry each claim.

Sources