Privacy-Preserving AI Audit Tools — OpenMined
OpenMined builds open-source software that enables external researchers to evaluate proprietary AI systems — answering questions about closed models without observing or copying them.
for AI Auditors OpenMined builds open source software which enables external researchers to evaluate proprietary AI systems With this software, external researchers can answer questions about proprietary AI systems without directly observing or acquiring a copy of the AI systems they are studying. The privacy-transparency tradeoff At the present moment, many powerful AI products and services remain unchecked by external parties. At the same time, the few external parties who do get to check these properties are often given too much access, exposing AI owners to numerous risks. This works…
saved by
related reading
- AVERI Pilot Report: The World’s First Double-Blind Evaluation of a Proprietary Language Model — AVERIaveri.org
- double-blind-evaluations-technical-report.pdfstorage.googleapis.com
- Oversight Assistants: Turning Compute into Understandingbounded-regret.ghost.io
- A Safe Path to Open Weights - Thinking Machines Labthinkingmachines.ai
- Auditing the singularitywuthejeff.com
- Attribution-Based Control for AI ⬩ OpenMinedopenmined.org
- Petri: An open-source auditing tool to accelerate AI safety researchalignment.anthropic.com
- The Open Anonymity Projectopenanonymity.ai
- Auditing language models for hidden objectives — LessWronglesswrong.com
- Auditing language models for hidden objectives \ Anthropicanthropic.com
- Petri: An open-source auditing tool to accelerate AI safety research \ Anthropicanthropic.com
- The Myth of unsafe Open Source AIflorianbrand.com