Claude API experienced elevated error rates across multiple models on February 25, 2026
Claude's status page reported elevated error rates affecting multiple models through their API endpoint at api.anthropic.com. The incident was tracked and resolved within approximately 31 minutes.
Timeline and Resolution
The incident followed this sequence:
- 17:15 UTC: Investigation began with initial status update
- 17:21 UTC: First update confirming continued investigation
- 17:21 UTC: Second identical update (likely system duplicate)
- 17:46 UTC: Incident marked as resolved
The status page indicates this was a specific incident affecting the Claude API service, not a general outage of all Anthropic services. The resolution came relatively quickly after detection.
Technical Context
When AI model APIs experience elevated error rates, this typically manifests as increased 5xx HTTP status codes, timeouts, or degraded response quality. For developers using Claude in automated workflows, such incidents can trigger retry logic, fallback mechanisms, or require temporary suspension of AI-dependent features.
The status page provides subscription options for incident notifications via email or SMS across international phone numbers, though the actual incident reporting was minimal on details about root cause or specific affected models.
📖 Read the full source: HN AI Agents
👀 See Also
Analysis of Claude Code's ~12K Token Forced System Prompt Reveals Priority Rules Overriding User Config
An analysis of Claude Code's injected ~12K token system prompt shows priority rules for song lyric bans, subagent delegation, and brevity that override user CLAUDE.md and memory files.
Claude 4.6 Opus Can Reproduce Linux's list.h From Minimal Input
A user demonstrated that Claude 4.6 Opus can generate a near-identical copy of Linux's list.h header file when given the first 43 lines as input with temperature set to 0, raising questions about GPL licensing implications for AI models trained on open-source code.
Coasty AI Agent Solves CAPTCHA Challenges Up to Level 6 Without Training
Coasty's Computer Using Agent (CUA) achieved 82% on the OSWorld benchmark, solving CAPTCHAs up to Level 6, browser popups, and cookie banners without specific training for 'I'm not a robot' challenges.
Anthropic Removes Claude Code from Pro Subscription for New Users in Test
Anthropic temporarily removed access to Claude Code from its $20/month Pro subscription plan for new users, changing website pricing pages and support documents before reversing the changes. The company described it as a 'small test of 2% of new prosumer signups.'