Microsoft AI Updates: August 1, 2026

1. Microsoft Announces Project Perception and Defender Prompt Injection Protection

Microsoft. In its July 2026 security update, Microsoft announced Project Perception, a coordinated multi-agent system in which red-team agents expose weaknesses, blue-team agents investigate threats, and green agents remediate issues in continuous loops. Microsoft also previewed prompt injection protection in Microsoft Defender that identifies and isolates emails containing malicious AI instructions before delivery. Source

2. Microsoft Research Releases Echoverse for Training Computer-Use Agents

Microsoft. Microsoft Research released Echoverse, a system for training computer-use agents in synthetic, evolving environments using a co-evolutionary loop of models, environments, and verifiers. A 9-billion-parameter model trained across twelve environments improved from 36.5% to 67.1% on the benchmark, and Microsoft released code, data, and graded test tasks for four environments on GitHub and Hugging Face. Source