Claude Service Incident: Elevated Errors Across Platforms
Service Disruption Details
On March 2, 2026, Claude experienced a service disruption affecting multiple platforms. The incident began with investigation at 11:49 UTC and was fully resolved by 15:47 UTC.
Affected Services
- claude.ai
- platform.claude.com (formerly console.anthropic.com)
- Claude API (api.anthropic.com)
- Claude Code
- Claude for Government
Timeline and Resolution
11:49 UTC: Investigation began
12:21 UTC: Initial findings indicated issues were related to claude.ai and login/logout paths, not the Claude API itself
13:22 UTC: Issue identified and fix implementation began
13:37 UTC: Discovery that some API methods were not working
14:05 UTC: Continued work on fix
15:25 UTC: Fix implemented, monitoring results
15:47 UTC: Incident resolved
The incident lasted approximately 4 hours from initial investigation to resolution. During this period, users experienced elevated errors when accessing Claude services through web interfaces and some API endpoints.
📖 Read the full source: HN AI Agents
👀 See Also
Claude Code v2.1.98 adds Vertex AI wizard, security fixes, and subprocess sandboxing
Claude Code v2.1.98 introduces an interactive Google Vertex AI setup wizard, adds subprocess sandboxing with PID namespace isolation on Linux, and fixes multiple security vulnerabilities including Bash permission bypasses and arbitrary code execution risks.
Claude Code v2.1.118 adds Vim visual mode, custom themes, and MCP improvements
Claude Code v2.1.118 introduces Vim visual mode with selection operators, custom theme management via /theme command, and multiple fixes for MCP OAuth authentication and plugin dependency resolution.
Claude Code Performance Regression Diagnosed: Configuration, Not Model Intelligence
Anthropic's postmortem reveals Claude Code's performance drop was caused by three product changes — default reasoning effort, session caching bug, and prompt-verbosity — not model degradation. Rollback restored performance.
AI Models Lack Self-Knowledge of Their Own Tools and UI
AI models like ChatGPT and Claude often provide incorrect or outdated information about their own features and interfaces, such as denying new slash commands exist or describing old UI versions, because they're trained on past snapshots while products evolve constantly.