AnthropicNews

Illustration: an hourglass with sand in the top bulb and a thin stream falling, in black and white.
Illustration

Claude spend-limit pauses and usage-page errors on Oct 6 and 7, resolved

If your org was refused on Oct 7 for hitting a spend limit it hadn't hit, Anthropic says that was its bug and it is fixed.

If Claude refused your requests on Oct 7 with a spend-limit error, you don't need to raise your limit. Anthropic says some organizations were "incorrectly paused," and the fix is in.

What broke, and when (ET)#

  • Opus 5.5 errors, Oct 6, 8:24 to 8:43 AM. Resolved.
  • Usage page and Admin API usage report errors, Oct 6, 8:00 to 9:40 PM. Resolved. Messages API requests were not affected.
  • Usage data errors again, from Oct 7, 9:25 AM. Mitigated at 5:44 PM and still listed as Monitoring: the last ~30 minutes of usage may show late. API requests were not affected.
  • Wrong spend-limit pauses, resolved Oct 7, 5:23 PM. Refused requests on the Claude API, Claude.ai, Claude Code, Claude Security and Claude Cowork.

Why it matters#

A spend-limit refusal looks like your own budget problem. Teams may have raised limits or paused jobs for no reason.

What to do#

  1. Check your org's spend limit in the Console and undo any raise made only because of Oct 7 errors.
  2. Rerun batch or scheduled jobs that failed that afternoon.
  3. Expect the newest ~30 minutes of usage to lag while the usage incident is in Monitoring.
  4. Subscribe at status.claude.com for updates.
  5. Claude Code users who hit 529 overloads can set CLAUDE_CODE_OVERLOADED_RETRY_MAX_DELAY_MS (2.1.296) for longer backoff.

Good to know#

  • Anthropic did not say how many organizations were paused or for how long.
  • Watch next: the usage incident moving from Monitoring to Resolved.

Sources#

Last checked: 2026-10-12

Sources

  1. status.claude.com
  2. status.claude.com
  3. status.claude.com
  4. status.claude.com

Something wrong? Tell the bot