All paid plans now get a global Chat/Work mode switcher in the desktop app (macOS and Windows), with unified Recents, Projects integration, and Work sessions syncing seamlessly across web, mobile, and desktop without re-explaining context.
Why it matters: Marketing and ops teams using ChatGPT for ongoing multi-step work now have genuine cross-device session continuity—the Chat/Work split makes it practical to run persistent agent-style task threads alongside fast ad hoc queries in the same tool.
The platform ships AgentTeams for multi-agent orchestration, Agentic Computer for sandboxed cloud execution, and a reusable enterprise skills layer—with internal pilots showing 15 coordinated agents handling 85% of developer support and compressing release cycles to one day.
Why it matters: Multi-agent infrastructure is entering the vendor evaluation stage; a third cloud-native agent stack alongside AWS and Azure means operators will start asking which platform to build on—and production benchmarks will matter more than demos.
The multi-model platform routes code vulnerability scanning across Microsoft, OpenAI, and Anthropic models, targeting broader enterprise access at lower cost than Mythos's limited preview—with a launch expected before end of July.
Why it matters: A more accessible AI security scanner opens the AI-native vulnerability conversation to enterprises priced out of Mythos's limited preview—and signals Microsoft competing on the same enterprise-security capability Anthropic has been building.