WPF advises NIST regarding synthetic content and data governance

WPF filed comments with the US National Institute of Standards and Technology regarding its draft, NIST Draft AI 100-4, Draft for Public Comment, Reducing Risks Posed by Synthetic Content: An Overview of Technical Approaches to Digital Content Transparency. 

WPF’s comments focused on 7 recommendations  ranging from technical to policy issues. One overarching recommendation was that NIST ensure that human rights were attended to in all of its plans. Additional recommendations include requesting that NIST attend to the risks of digital exhaust in metadata, ensure that biometric data is included in the guidance, among other recommendations, including:

  • Utilize normative codes of conduct for Standards Development Orgnizations in its work, such as the code of ethics that ISO follows that provides for non-dominance and openness.
  • Attend to the risks of digital exhaust in metadata, including the risk of onward sharing of metadata that has governance and human rights implications.
  • Focus on evaluating governance and privacy methods for synthetic content detection, noting that there are difficult trade-offs involved, some of which have the potential to negatively impact the lives and privacy of vulnerable groups, among others.
  • Include biometric data as an area of research in its synthetic content metadata work. WPF noted that the exclusion of biometric data in the draft was an oversight, and should be corrected.
  • Ensure that metrology – or measuring the measures – becomes an integral part of the NIST work. Tools and metrics developed under the NIST AISIC need to be fit for purpose and reliable, and provably achieve their AI measurement goals. (Fairness, transparency, fit, etc.)

WPF is part of the NIST AI Safety Institute Consortium.

Documents: