Hi AI ethics enthusiasts,
Today, I'm excited to share with you a glimpse into the future of AI governance in the public sector.
My team and I recently submitted recommendations to the US Office of Management and Budget (OMB) on responsible AI procurement in government. These insights, drawn from our ongoing research on AI in public administration, could shape how our government interacts with AI technologies.
I will walk you through some key points and why they matter. The full recommendations are here.
For dessert, an AI-generated take on this post in the style of infographic meets official government seal meets cyberpunk!
The Context: OMB's Quest for Responsible AI
The US Office of Management and Budget (OMB), a critical player in shaping US government policies, is currently developing guidance on AI use in government. As part of this process, they issued a Request for Information (RFI) on responsible AI procurement.
We responded to the RFI. Our insights are based on an ongoing research project on AI in public administration. Stay tuned for the paper about that!
The team behind these recommendations:
Nari Johnson, Hoda Heidari, Motahhare Eslami, and Harrison Leon
From Pitt Cyber at the University of Pittsburgh.:
Beth Schwanke, Jessica Kuntz, and Elise Silva.
Extra shout out to Nari and Jess, who spearheaded the writing of this document!
Selected Recommendations
1. Pricing as a barrier to responsible AI
We urge OMB to limit federal agency use of the lowest price technically acceptable (LPTA) process in AI procurement. We would argue that ensuring the selected service provides sufficient transparency to screen for bias and other harms proactively may be a higher priority than price.
2. Acquisition beyond formal procurement
OMB should consider how best to apply standards and regulations developed for AI procurements to other forms of acquisition other than formal procurement, such as purchasing cards, research collaborations, donations, piggyback contracts, and other contracts under a set dollar amount exempt from full procurement procedures.
3. Identify AI
Each agency's procurement officials should be trained to identify and ask vendors if they plan to use AI to complete the scope of work.
4. AI disclosure
Agencies should include a standard contract clause requiring vendors to notify the agencies when they incorporate new AI features into existing procurements.
5. AI opt out
When appropriate, agencies should include a standard contract clause requiring vendors to implement and provide "opt-out" of AI functionalities introduced after an initial purchase.
What's Next?
Our recommendations are now with the OMB, and they are publicly available here. We're continuing our research on AI procurement in public administration, and I'll be sure to keep you updated on any developments.
I'd love to hear your thoughts on these recommendations. Do you think they go far enough? Are there other aspects of government AI use that you think need addressing?
Dessert
An AI-generated take on this post in the style of infographic meets official government seal meets cyberpunk!
Ready for More?
Check out our comprehensive resources, workshops, and consulting services at www.techbetter.ai, and follow us on LinkedIn: Ravit's page, TechBetter's page.