🚀 Agentic Momentum: Big Week for Copilot & Enterprise AI

[ebndev]

We’re shipping a wave of features that level up how your org adopts, governs, and collaborates with AI on GitHub. From enterprise-grade control to seamless delegation in more IDEs, to smarter, safer automated contributions—Copilot just got a lot more powerful.

🛡️ Enterprise AI controls & the Agent Control Plane (public preview)

Enterprise Cloud admins now have a dedicated home to govern AI systems:

• Unified “AI controls” view for policies, configuration, and oversight.
• Manage today’s handful of agents—and tomorrow’s fleet.
• See 24h agent session activity with filters + deep links.
• New audit log fields surface exactly what agents did and on whose behalf.
• Coming soon: fine-grained permission (FGP) to decentralize AI admin without broad enterprise access.
• Set enterprise MCP allowlists to safely govern tool connections.

This is your foundation for safe, scalable agent adoption.

Visit the changelog for more information.

🧠 Custom agents, enterprise-ready

Standardize rollout:

• Pick one org as the source of truth for .github/agents/*.md.
• Use a 1‑click push rule to protect agent definitions.
• Let teams experiment while keeping guardrails in place.
• Select enterprise-approved custom agents when assigning Copilot.

✍️ Delegate coding tasks from JetBrains IDEs

Public preview: You can now hand work to Copilot coding agent directly inside JetBrains IDEs.

1. Ensure your Copilot plan & policies are enabled.
2. Open Copilot Chat (ask/agent mode).
3. Describe a task → hit “Delegate to Coding Agent.”
4. Track progress via job cards and sidebar, stay focused on your flow.

Visit the changelog for more information.

🔐 Automatic security & quality validation (no extra setup)

When Copilot coding agent writes code, it now:

• Scans for vulnerabilities with CodeQL.
• Checks new dependencies against the Advisory Database.
• Runs secret scanning.
• Performs a code quality review.

It attempts fixes before asking for your review—all included without needing GHAS.

Visit the changelog for more information.

🗨️ Inline collaboration: change any pull request with @copilot

Ask for edits directly in a PR comment. Mention @copilot:

• Copilot opens a new PR targeting the original PR’s branch.
• Your original PR stays untouched until you merge its proposed changes.
• Parent/child PR relationships are visually clear.

Fast, focused iteration—still under your control.

---

Why this matters

You’re getting tighter governance, more flexible adoption paths, broader IDE support, and stronger built-in safeguards—all while keeping humans firmly in charge of review and acceptance.

Jump in

• Explore the new AI controls area (Enterprise Cloud).
• Try delegating a JetBrains task you’d normally do manually.
• Drop a comment with @copilot on an in-progress PR and see the workflow.
• Watch your agent activity + audit logs to build trust internally.
• Start drafting a custom role for upcoming AI admin FGPs.

Join the conversation below and tell us what you want next.

Let’s build faster—safely—together. 💙

[adcloeshelle]

@ebndev - I’m super excited about all the announcements from GitHub Universe today! One announcement I saw from https://github.blog/news-insights/company-news/welcome-home-agents/ was "New branch controls that give you granular oversight over when to run CI and other checks for agent-created code.”

Is there any further information or documentation on these new branch controls? I was trying to navigate the website / look for more details but couldn’t find it.

Some things that would really help our use case:

• The ability to set different minimum approvals for PRs created by the Coding Agent
• The ability to set default copilot setup instructions across the entire organization