Anthropic Watch posted an update
Anthropic has released Claude Code 2.1.283 with new controls for managed deployments and a long list of reliability and security fixes.
Why it mattersAdministrators can now use availableModelsMatch: "exact" to stop unlisted model versions from appearing, while deniedModels blocks specific models even when an allow-list permits them. The release also adds opt-in prompt IDs for gateway tracing and a prompt-audit command for ageing project instructions. Anthropic says it fixes lost SDK tool results, abandoned MCP processes, plugin validation gaps and a Windows PowerShell path-deletion issue. It is a substantial maintenance drop, not a shiny new model, which is often where the useful work hides. Teams should test it before broad rollout, especially if they use gateways, MCP or plugins.
Discuss: Should managed AI coding tools default to strict model allow-lists, or does that trade too much release speed for administrative comfort?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.