Claude Code
AI Agent Reliability Report
According to StupidLLM's incident database, Claude Code has 8 documented failures with an average severity of 6.4/10 — including 3 critical and 2 high-severity incidents.
6.4
Avg Severity /10
8
Total Incidents
3
Critical
2
High
Failure Modes
Destructive Action3
Security Vulnerability2
Other2
Hallucination1
Root Causes
Confidence Miscalibration2
Instruction Misunderstanding1
Training Data Gap1
Other1
Scope Misunderstanding1
Tool Misuse1
Context Overflow1
Worst Documented Incident
STUPID-2026-000310.0criticalVerified
Claude Code ran rm -rf on test fixtures thinking they were temp files
All Claude Code Incidents
STUPID-2026-000310.0criticalVerified
Claude Code ran rm -rf on test fixtures thinking they were temp files
STUPID-2026-004110.0criticalVerified
Claude Code wiped DataTalks.Club's production infrastructure — 2.5 years of course data — during an AWS migration
STUPID-2026-004210.0criticalVerified
Claude Code ran rm -rf from the filesystem root, destroying a developer's home directory (GitHub #10077)
STUPID-2026-00197.5highVerified
Claude Opus 4.5 leaked API key in console logs during YouTube scraper build
STUPID-2026-00247.5highVerified
Claude Code MCP trust boundary failures allow workspace privilege escalation
STUPID-2026-00452.2lowVerified
Anthropic admitted a month of Claude Code degradation: lost context, repeated steps, burned usage
STUPID-2026-00552.2lowVerified
Uber burned its entire annual AI coding budget in ~4 months after rolling out Claude Code to 5,000 engineers
STUPID-2026-00082.1lowVerified