Updates
Featured
What the OpenAI-Hugging Face Hack Really Tells Us About AI Danger
A new method for studying near-verbatim memorization in language models that has implications for access requirements in frontier AI auditing
AVERI’s first endorsement of specific legislation
Former OpenAI policy chief creates nonprofit institute, calls for independent safety audits of frontier AI models
A comprehensive framework for independent evaluation of frontier AI systems, mapping access requirements to systemic risks.
Without reliable measurement of AI systems, we cannot conclude a measured system is safe
The BenchRisk workflow allows for comparison between benchmarks; as an open-source tool, it also facilitates the identification and sharing of risks and their mitigations.