As artificial intelligence becomes more ingrained in all aspects of technology, agencies and vendors alike must consider new factors when it comes to buying these capabilities.
In a new memo released today, the Office of Management and Budget outlined these considerations in three buckets:
- Proactively manage AI risk and performance
- Promote a competitive AI marketplace
- Create structures to govern and manage business processes related to acquiring AI in acquisition
“This new memo provides agencies with the tools and information they need as they acquire AI capturing his promise while managing its risks,” said Jason Miller, OMB’s deputy director for management, in a call with reporters. “The AI landscape is rapidly shifting and today’s guidance is an initial means for responsible AI acquisition by federal agencies.”
Miller said the memo both fulfills one of the requirements in President Joe Biden’s executive order on responsible AI from last October and tries to get ahead of the learning curve as more and more agencies go deeper into use AI.
In fact, OMB is expected to release an updated governmentwide inventory of AI uses cases on or around Dec. 16. The 2023 inventory showed there were more than 700 AI use cases, but the expectation is that number could easily double or triple given the excitement around generative AI and expanded use of predictive AI. A Government Accountability Office report from last December found 1,200 existing use cases, but most of the reported AI use cases are “in the planning phase, and not yet in production.
This is why OMB’s 36-page memo is trying to address some current and emerging questions around buying AI.
“Acquisition of AI presents some novel challenges, even though in some respects, much of the acquisition is similar to purchases of software without AI,” Miller said. “The due diligence that informs our buying process is a critical moment to ensure we’re setting ourselves up for success, that federal agencies are setting themselves up for success consistent with our expectations of agencies that when they are using AI to improve mission delivery for American citizens, they’re doing so in a way that is consistent.”
To ensure that agencies are applying the right rigor to the AI tools that are aimed at improving agency missions, OMB defined what type of AI falls under this memo and what type does not.
“Data systems, software, applications, tools or utilities established primarily for the purpose of researching, developing or implementing artificial intelligence technology, as well as data systems, software, applications, tools or utilities where an AI capability is integrated into another system or agency business process, operational activity or technology system,” OMB wrote. “The term excludes, however, any common commercial product within which artificial intelligence is embedded, such as a word processor or map navigation system.”
Avoiding AI vendor lock-in
OMB also detailed what types of AI the memo doesn’t apply to, such as “evaluations of particular AI applications because the AI provider is the target or potential target of a regulatory enforcement, law enforcement or national security action;” or “agencies’ acquisition of AI to carry out basic, applied or experimental research except where the purpose of such research is to develop particular AI applications within the agency.”
Once agencies have decided these rules apply to the AI tools or capabilities they are buying, OMB says another key factor to consider when trying to ensure a competitive market.
OMB’s Miller said the memo focused on concepts like interoperability, transparency, data portability and intellectual property rights.
“We want to ensure choice and diversification in the federal marketplace, to demonstrate our agility for our vendors can continue to innovate to improve AI use,” he said. “We want to ensure we’re always, always getting the best price for U.S. taxpayers on the products and services that we are acquiring.”
For example, agencies should “develop an approach to IP that considers what rights and deliverables are necessary for the agency to successfully accomplish its mission, protects federal information used by vendors in the development and operation of AI systems and services for the federal government, considers the exploration of open-source development practices of AI code, avoids vendor lock-in and avoids unnecessary costs.”
CXO coordination is essential
This could include negotiating the appropriate scope of licensing rights and other rights as well as creating a process and timeline for delivery of components needed to operate and monitor the AI system, including as appropriate: data; inputs to the development, testing and operation process.
The third bucket OMB is asking agencies to focus on is managing their business and acquisition processes.
Miller said this means creating a “tight loop” between leaders inside the agency and across the government.
The memo both calls for increased cross-agency coordination of the different councils like chief information officer, chief acquisition officer, chief data officer and chief AI officer as well as efforts inside agencies to bring together these C-suite leaders.
“We are sharing and we are reviewing the use case inventory together. We’re bringing cohorts of people together. We’re all learning fast here. And by bringing folks together and finding that multiple teams are trying to solve the exact same thing so by bringing them together, we are learning from a pilot that is already further down the path than somebody else’s, we are really trying to focus on moving quickly,” said a senior administration official. “We are making sure that we are sharing those key learnings and people are sharing non-sensitive procurement and acquisition information … so that we can all learn from each other. I think this is one of the most powerful ways that we’re going to go fast and not duplicate efforts by buying the same thing 400 times.”
OMB has been working on this memo since April when it issued a request for information and held an industry roundtable as part of it effort to gather feedback on how best to address challenges of acquiring AI.
A second senior administration official said OMB received more than 60 comments, which equaled more than 430 pages, from the RFI.
“We have coordinated this work consistently with all of those council groups. Then we have been working on the specific guidance individually with each agency. This is no surprise to them,” said the first administration official. “They have been partners with us because we have wanted to hear from the people on the ground who are doing the work to make sure that we are writing guidance that both achieves the administration’s goals and also is actionable, so that the agencies have clear comprehension and understanding so they can meet the deadlines and that we can continue to manage this effectively.”
Quinn Anex-Ries, a senior policy analyst at the Center for Democracy & Technology, said in a statement that OMB’s new memo provides agencies with a clear roadmap to ensure the AI they buy protects the rights and safety of the public.
“As federal spending on AI increases, this guidance is not only an important framework for protecting the public from potential harms but for encouraging responsible AI practices throughout the AI marketplace more broadly,” Anex-Ries said.
The post OMB gives agencies new considerations when buying AI first appeared on Federal News Network.
