Open source has a growing problem with LLM generated issues (github.com)

Maintainers of open-source projects (notably the opencontainers/runc team) are seeing a rising number of pull requests and bug reports that appear to be generated by large language models, and they’re debating how to handle them. The suggested immediate steps include documenting a policy in CONTRIBUTING.md, treating LLM-generated issues as spam to be closed because their descriptions often contain extraneous or incorrect details, and treating LLM-generated code differently: accept only when the submitter can explain and defend the change in their own words. The discussion cites concrete examples (#4982, #4972 for issues; #4940, #4939 for patches) and raises the minority legal view that LLM-produced code may fail to meet DCO (Developer Certificate of Origin) requirements. This matters to the AI/ML and OSS communities because LLM-assisted submissions change triage workload, signal-to-noise, and trust in issue reports and patches. Practically, maintainers may need new norms and tooling—explicit contribution policies, prompts or provenance disclosure, verification steps, and stricter review gates—to ensure reproducibility, accountability, and legal compliance. The conversation highlights a clear split between treating generated issue noise as spam versus allowing assisted code only when the human contributor demonstrates true understanding.