Misuse

Question: Is AI use causing harm or violating community norms and expectations including the Code of Conduct.

This metric is part of the AI Alignment - Community Governed Use metrics model.

Overview

Capture harm and disruption that predates or falls outside written policy, so communities without an AI policy or that which violates the Code of Conduct. Other potentials:

• License laundering — AI regenerating code from training data in a way that strips out copyleft reciprocity requirements
• Wrongful accusation — contributors who did not use AI being wrongly accused of undisclosed use; harm produced by retaliation and other motivations.

Want to Know More?

Data Collection Strategies

• Self-disclosed AI use by human contributors: depends on policy and culture; no technical enforcement
• Whether a contribution came from any LLM at all (no disclosure): unknown; text-based AI detectors are unreliable
• Which specific model produced an undisclosed contribution: unknown; no reliable signal currently exists
• Whether an approved model was used vs. an unapproved one in undisclosed contributions: unknown; depends on the above
• License 'laundering' incidents: maintainer or legal self-report; no reliable automated signal currently exists
• Wrongful accusation incidents: maintainer self-report, code of conduct records; no reliable automated signal currently exists.

Note: The "unknown" entries are where the CHAOSS and adjacent projects can build, through tooling, partnerships, or formally surfacing the gap as unsolved.

Filters

• Code of conduct category
• Risk type (safety, security, reputation, privacy)

Visualizations

• A bar chart or timeline that shows the number and severity of misuse events over time.

References

• Contributors Covenant 3.0
• CHAOSS safety metric

Contributors

• Emma Irwin
• Adrian Edwards
• Justin Wheeler
• Coraline Ada Ehmke
• Omkar Petkar
• Andrew Nesbitt

Add your name if you have contributed

Additional Information

To edit this metric please submit a Change Request here: (replace with correct link)

To reference this metric in software or publications please use this stable URL: (replace with permalink)