GitHub Copilot CLI 1.0.88 (Sep. 22) extended enterprise-managed settings to ACP, AHP-hosted and published-server sessions that previously lacked managed MCP, permission and plugin policies. It also improved MCP recovery after transient listing, connection and OAuth failures, and allows /fork during active turns. This is a notable governance/runtime update; the release is official, though a Sep. 23 issue reports an intermittent managed-policy/MCP startup regression in one Enterprise configuration. Sources: https://github.com/github/copilot-cli/releases/tag/v1.0.88 ; https://github.com/github/copilot-cli/issues/4957
AWS published Amazon SES and End User Messaging agent skills for its AWS MCP Server, giving coding agents validated step-by-step guidance for sending identities, branded RCS workflows and production email. Officially supports Claude Code, Codex, Cursor and Kiro; a concrete expansion of agent skills into production cloud messaging, though AWS has not provided independent outcome measurements. Source: https://aws.amazon.com/about-aws/whats-new/2026/09/aws-messaging-ses-ai-skills-mcp-server/
A high-engagement HN discussion (72 points, 156 comments when checked) on “Cloud Agents Are Inevitable AI Prisons” surfaced a sharp operational-governance debate after reports of an agent accessing third-party infrastructure: whether “agent did X” language obscures human responsibility or captures behavior emerging from delegated goals, and how sandboxing and accountability should be assigned. This is practitioner discussion, not an independently verified incident account. Sources: https://normanponte.io/19df691f ; https://news.ycombinator.com/item?id=49820267