Signal, not noise: five primary-source reads from September 10, each with a two-line note on why it matters for builders.
Introducing the Agents API
OpenAI now offers the Codex harness as a managed public-beta API with hosted sandboxes or self-hosted execution and no extra fee beyond tokens and tools. Start new agent work on the managed harness and keep your effort on tools, knowledge, and workflows.
Read the full story at OpenAI (September 10, 2026).
Detecting and countering misuse of AI: September 2026
Anthropic reports eight months of disrupted Claude misuse across cyber, surveillance, fraud, biology, weapons, and distillation, with accounts banned and safeguards strengthened. Treat the case studies as a checklist for logging, abuse detection, and disclosure before your agent ships.
Read the full story at Anthropic (September 10, 2026).
Measuring AI capabilities in intelligence targeting and conventional weapons
Anthropic's Frontier Red Team adds evaluations for tactical targeting and conventional-weapons development, showing steady progress and new classifiers to block such misuse. Gate dual-use benchmarks behind review and add classifiers before exposing planning or engineering helpers.
Read the full story at Anthropic (September 10, 2026).
Use Cloudflare Containers with Codex via the OpenAI Agents API
Cloudflare Containers can serve as self-hosted execution for the Agents API, with a reference Worker that keeps a container per Codex session and shuts down when idle. Copy the per-session container pattern when you need customer-controlled execution with network and data limits.
Read the full story at Cloudflare (September 10, 2026).
AI Scan for pull request APIs in public preview
GitHub adds REST endpoints to manage AI Scan enablement for pull requests at organization and repository level for Advanced Security customers. Roll out AI-powered detections programmatically per repository instead of clicking through settings.
Read the full story at GitHub (September 10, 2026).