Slack launched Slack Code on August 20, embedding Claude Code, Devin, GitHub Copilot, and Vercel's agent into dedicated coding channels available on any Slack plan.

• Teams can plan, steer, and review AI-generated code inside a shared channel rather than through individual terminal sessions, shifting agentic coding from a solo to a collaborative workflow.
• ChatGPT integration is on the roadmap; customers must hold their own licenses to each partner agent, so this is an orchestration layer, not a bundled service.

Salesforce expanded Headless 360 across its platform today, adding Model Context Protocol servers, Data 360 functions, Slack integrations, and more than 100 reusable Skills so authorized agents can traverse the full Salesforce estate without custom integrations.

• The MCP Server at the center lets any compliant AI agent discover and invoke business functions across Sales Cloud, Service Cloud, and other Salesforce products through a single protocol.
• Developer tooling and Slack hooks are included, suggesting Salesforce intends Headless 360 to serve both internal Agentforce agents and third-party agents routed through Slack Code.

JetBrains shipped Rider 2026.2.1 with a bundled skill that gives AI agents direct access to Rider's semantic refactoring engine, dropping median task time from 157.9 seconds to 26.6 seconds.

• Cost per refactor task fell from $0.52 to $0.19, and build invocations across a 15-task evaluation collapsed from 163 to 3, because agents no longer resort to text-level edits that break the build.
• The skill works with Claude Code and Codex today; any agent that previously struggled with C# rename-or-extract tasks on a .NET codebase gets the improvement immediately on update.

Microsoft disclosed and patched CVE-2026-69836, a maximum-severity remote code execution vulnerability in Entra ID caused by deserialization of untrusted data, confirmed as already exploited in attacks.

• The flaw is cloud-side, meaning enterprise customers could not patch it themselves and had to wait for Microsoft's remediation, a structural risk in cloud-only identity platforms.
• Entra ID is the authentication backbone for Microsoft 365, Azure, and Dynamics, so a pre-patch exploit window directly threatens every tenant relying on conditional access policies.

OpenAI confirmed Microsoft and Databricks as early customers testing Private Safety Processing, its zero-data-retention safety system that detects misuse patterns across sessions without retaining prompts or exposing content to OpenAI staff.

• The system detects risk across multi-session interactions rather than per-message, addressing a compliance gap that previously forced enterprise buyers to choose between frontier model capability and data-retention restrictions.
• A technical white paper and broader rollout are planned for September, meaning enterprises negotiating API contracts now have a concrete timeline to evaluate against Anthropic's competing 30-day log requirement.