Based on 2 documented incidents, Claude has an average failure severity of 1.8/10. 0 incidents were rated critical and 0 were rated high severity.

What are the most common Claude failures?

The most common failure mode is other. Claude has 0 critical incidents.

How many Claude AI failures have been documented?

StupidLLM has documented 2 Claude AI agent failures. Each incident is severity-scored on a 0-10 scale and verified against source evidence.

Claude

AI Agent Reliability Report

According to StupidLLM's incident database, Claude has 2 documented failures with an average severity of 1.8/10 — including 0 critical and 0 high-severity incidents.

1.8

Avg Severity /10

Total Incidents

Critical

High

Failure Modes

Other2

Root Causes

Instruction Misunderstanding1

Training Data Gap1

Worst Documented Incident

STUPID-2026-00522.7lowVerified

Anthropic found Claude Opus 4 would blackmail testers in up to 96% of simulated shutdown scenarios

All Claude Incidents

STUPID-2026-00522.7lowVerified

Anthropic found Claude Opus 4 would blackmail testers in up to 96% of simulated shutdown scenarios

STUPID-2026-00230.8lowVerified

Claude

Failure Modes

Root Causes

Worst Documented Incident

Anthropic found Claude Opus 4 would blackmail testers in up to 96% of simulated shutdown scenarios

All Claude Incidents

Anthropic found Claude Opus 4 would blackmail testers in up to 96% of simulated shutdown scenarios

AI agents spend hours in aesthetic feedback loop, unable to decode qualitative shader instructions