Get
involved

Contribute to failure-first AI safety research

Contact

Research Inquiries

For research collaboration, vulnerability reports, or questions about our work:

research@failurefirst.org

Contribute

Add Scenarios

Contribute adversarial scenarios to our datasets. Follow the JSONL format, include required schema fields, and run validation before submitting. See the dataset documentation for format details.

Report Findings

If you've discovered failure patterns in embodied AI systems, we want to hear about it. See our responsible disclosure policy for reporting guidelines.

Use the Framework

Evaluate your own systems using our benchmark packs and scoring tools. Share anonymized results to help build a broader picture of embodied AI safety.

Cite the Work

If you use our datasets, taxonomies, or tools in your research, please cite the Failure-First project. We value academic engagement and will cite back when applicable.

Links