STUPID-2026-0037
AI 'slop' vulnerability reports flooded curl until it killed its bug bounty
Why 5/10
Bogus AI reports drained a volunteer security team until curl ended its bug bounty. The cost is maintainer time and a closed programme, not data.
- Scope
- 1/4
- A few files, one session, one credential, or a small cost
- Recoverability
- 1/3
- Fully recovered, with effort
- Did it happen
- 2/2
- Happened in real use
- Beyond the user
- 1/1
- Production systems, customers or third parties, or credentials and sensitive personal data
The score is the sum of these four factors. How each factor is scored · scored 2026-10-29
Instruction given
Generate a security vulnerability report to submit to curl's bug bounty.
Expected behavior
Only report real, reproducible vulnerabilities in code that actually exists.
Actual behavior
LLMs produced long, confident, fabricated vulnerability reports — one 'HTTP/3 stream dependency cycle exploit' came complete with GDB sessions and register dumps referencing a function that does not exist in curl. Roughly one in five 2025 submissions was outright AI slop.
Damage
The confirmed-vulnerability rate fell from above 15% to below 5%. Each bogus report still ate hours from curl's seven-person volunteer security team. Maintainer Daniel Stenberg ended the curl bug bounty entirely at the end of January 2026 to remove the payout incentive.
Full write-up
curl's founder Daniel Stenberg shut down the project's long-running HackerOne bug bounty at the end of January 2026 after being overwhelmed by AI-generated 'slop' — long, confident, and often entirely fabricated vulnerability reports produced by LLMs. One especially egregious submission described an 'HTTP/3 stream dependency cycle exploit' complete with GDB sessions and register dumps, all referencing a function that does not exist anywhere in curl. As AI slop climbed to roughly one in five submissions, curl's confirmed-vulnerability rate collapsed from above 15% to below 5%, while each bogus report still consumed hours from a seven-person volunteer team. Stenberg described it as AI 'DDoSing' open source: hallucinated bug reports that cost real maintainer time and ultimately killed the incentive structure of a well-run security program.
Classification
- Agent
- Multiple LLMs
- Failure mode
- Hallucination
- Root cause
- Confidence Miscalibration
- Domain
- Backend
- MIT AI Risk Repository (our mapping)
- 3.2 Pollution of information ecosystem and loss of consensus reality taxonomy
- Source
- News Report
Related incidents
Cite this record
StupidLLM. "STUPID-2026-0037: AI 'slop' vulnerability reports flooded curl until it killed its bug bounty." AI Agent Incident Database, 2026. https://www.stupidllm.com/incident/STUPID-2026-0037/ (accessed 2026-10-29). CC BY 4.0.
← STUPID-2026-0036By IDSTUPID-2026-0038 →
Get told when an agent breaks something
We document AI agent failures daily, severity-scored against a published scale. Each Monday, if any scored 7 or above that week — deleted data, leaked secrets, broken production — you get one email with the sources. When nothing did, you get nothing.
This database is callable over MCP — query it from inside your agent.