Toward Trustworthy AI Development - Heidy Khlaaf co-authored a new multi-stakeholder report
Heidy Khlaaf, Senior Consultant at Adelard, joined the international team of academic researchers, industry scientists, policy experts and other stakeholders involved in the development of AI to co-author “Toward Trustworthy AI Development: Mechanisms for Supporting Verifiable Claims” report.
The growing impact of AI systems unveiled the fact that current regulations and norms aiming to ensure responsible AI development are insufficient. This joint report addresses current gaps preventing effective assessment of verifiability of AI developers’ claims and suggests 10 mechanisms for how AI developers can make more verifiable claims in three areas: institutional, software and hardware.
The report also refers to CAE (Claims, Arguments, and Evidence) framework, which ensures claims verifiability through its other 2 key elements - arguments and evidence, and is now being increasingly adopted to the safety analysis of AI systems work.