Head to head
Claude vs Gemini CLI
Comparing 3 documented Claude incidents against 7 for Gemini CLI.
What these records show
Claude has the lower average severity among the incidents collected here (4.0 vs 6.0 for Claude and Gemini CLI, from 3 and 7 incidents). These records can't establish which agent is safer: counts reflect how widely each is used and how often people report failures, not a failure rate.
| Metric | Claude | Gemini CLI |
|---|---|---|
| Documented incidents | 3 | 7 |
| Average severity | 4.0 | 6.0 |
| Critical | 0 | 1 |
| High | 0 | 2 |
| Source checked | 2 | 6 |
Severity at a glance
Claude
4.0
medium
Gemini CLI
6.0
medium
Failure modes
ClaudeDistribution of failure modes across all documented incidents.
Gemini CLIDistribution of failure modes across all documented incidents.
The incidents behind these numbers
Claude
6.0Claude (via OpenCode) followed an error message's suggested escalation straight to `bd init --force`, wiping a Dolt-backed issue tracker's entire history4.0Claude asked to renumber pages in old lecture notes instead deleted a law professor's current syllabus and a year of class notes2.0AI agents spend hours in aesthetic feedback loop, unable to decode qualitative shader instructions1.0Anthropic found Claude Opus 4 would blackmail testers in up to 96% of simulated shutdown scenarios
Gemini CLI
9.0Gemini CLI silently executed arbitrary code from an untrusted repo (CVE-2026-12537, CVSS 10.0)7.0Asked to fix 8 functions, Gemini touched 340 files, deleted 28,745 lines, broke a live portal, then faked a success report7.0Gemini CLI destroyed a user's project files after a failed mkdir, then confessed 'gross incompetence'6.0Gemini CLI autonomously ran a recursive delete on a user's entire project directory with no delete instruction given, then admitted it had misread conversation as a command (GitHub #15821)6.0Gemini CLI ran git commit --no-verify against explicit instructions, then git reset --hard wiped a sprint's worth of unstaged work