Fable 5 included subscription access window closes today: Anthropic planned June 23 removal to usage credits, but global suspension since June 12 still blocks all access
The June 9 launch post promised Fable 5 at no extra cost on Pro, Max, Team, and seat-based Enterprise plans through June 22, with removal from those plans on June 23 and usage credits required after that. The June 12 export-control suspension disabled Fable 5 and Mythos 5 worldwide before most teams finished piloting, so this calendar milestone is mostly about routing and budget planning, not a live model you can call today. If you pinned fable-5 in configs before the suspension, keep Opus 4.8 as fallback and treat the June 23 credit shift as irrelevant until Anthropic restores access. When Fable returns, re-run cost-per-completed-task evals against Opus fast mode before promoting it back to default. Routing playbook: /blog/claude-fable-5-frontier-models-for-agent-builders.
Claude Code ships Artifacts in beta: Team and Enterprise sessions publish live, org-private review pages that update in place at a claude.ai URL
Artifacts turn terminal output into a single self-contained HTML page on claude.ai that republishes to the same URL as the session progresses, with version history and org-only sharing from the page header. Claude builds from session context (codebase, connectors, conversation), so PR walkthroughs, incident timelines, and dashboards do not need a separate export step. Pages run under a strict CSP with no external fetch and no backend, so they are review captures, not hosted apps. Requires Team or Enterprise, /login to claude.ai (not API-key auth), and Anthropic API routing (not Bedrock or Vertex). If your reviewers still get screenshots in Slack, pilot one incident or PR artifact this sprint and set retention in admin settings before wider rollout. Governance context: /blog/governing-agent-autonomy-auto-review-and-pre-push-review.
Claude Code v2.1.183 tightens auto-mode safety: blocks destructive git resets, agent-amend commits, and infrastructure destroy unless you asked for them
Auto mode now refuses git reset --hard, git checkout -- ., git clean -fd, and git stash drop when you did not ask to discard local work; blocks git commit --amend on commits the agent did not make this session; and blocks terraform destroy, pulumi destroy, and cdk destroy unless you named the stack. The same release warns on stderr when a requested model is deprecated or auto-upgraded (including models set in agent frontmatter), fixes WebSearch returning empty results in subagents, and stops scheduled task and webhook deliveries from being treated as keyboard input that could approve pending actions in auto mode. If you run auto mode for throughput, upgrade to v2.1.183+ and keep secret scanning at commit time because these guards are convenience, not a compliance boundary. Guardrails guide: /blog/governing-agent-autonomy-auto-review-and-pre-push-review.