Curated AI news from the last 24 hours — models, agents, enterprise, and developer tools.
OpenAI released the Agents API in public beta, letting developers build and run cloud-based agents backed by the Codex harness while OpenAI manages orchestration, long-running sessions, and context compaction. Developers supply tools and choose an execution environment, with no additional platform fees beyond standard token and tool usage.
openai.com ↗xAI expanded Grok Bot, its persistent AI agent platform, to enterprise customers with new access, network, and audit controls for governing bots at scale. Each Grok Bot runs autonomous, long-running tasks in an isolated cloud environment, and xAI says thousands of organizations, including Legora, Supermicro, and ServiceTitan, have adopted it since launch, with the heaviest usage outside engineering teams.
x.ai ↗GitHub's weekly Copilot roundup added Jira integration to the Copilot app and introduced adaptive model orchestration via Project HydraFusion in Copilot CLI. The release also brought new agent automation to Copilot in Visual Studio Code and expanded enterprise controls for Copilot in JetBrains IDEs.
github.blog ↗GitHub launched a public preview of AI Scan, a feature that applies AI-driven analysis to pull request APIs to surface issues before merge. It extends GitHub's code review tooling with automated scanning aimed at catching problems earlier in the review cycle.
github.blog ↗GitHub added settings that let organizations tune how Copilot's automatic model selection balances cost against output quality, rather than relying solely on GitHub's default routing. The change gives admins more direct control over spend as Copilot usage scales across teams.
github.blog ↗GitHub updated Copilot code review with auto-resolution capabilities that let the agent close out certain flagged issues automatically, alongside improved analysis quality. The update is aimed at reducing manual triage overhead for reviewers using Copilot-assisted code review.
github.blog ↗GitHub expanded its Copilot usage metrics and admin dashboards to include activity from VS Code Agents, giving organizations visibility into agent-driven work alongside standard Copilot usage. The addition rounds out reporting for teams tracking AI-assisted development at scale.
github.blog ↗GitHub deprecated the MAI-Code-1-Flash model across all Copilot experiences, removing it from the set of selectable models. Users of the model are directed to migrate to currently supported alternatives in Copilot's model picker.
github.blog ↗