The White House Office of Management and Budget (OMB) has released two key AI policy documents with marching orders for Federal agencies this year, with looming deadlines in both November and December.
Samantha Hubner, the acting director of AI policy at the Office of Management and Budget (OMB), joined a virtual fireside chat with global tech trade association ITI on Thursday to discuss the two policies and what the deadlines mean for Federal agencies.
The first AI policy that OMB published in March of this year is M-24-10, which sets requirements to promote the safe, secure, and responsible use of AI by Federal agencies. The memo delivered a core component of the administration’s October 2023 AI executive order (EO).
OMB released the second AI policy earlier this month, M-24-18, which directs Federal agencies to improve their capacity for the responsible acquisition of AI.
“When you put this in that broader context of the administration and all the work that’s been happening on Federal use of AI, I think you can, of course, first point to the executive order and the way that that directed forth specific deadlines for OMB’s first AI memo, M-24-10,” Hubner said.
“A big piece of this guidance [M-24-18] is just ensuring that we take those same principles and ensure that those contracts that are for AI systems that they align with what we have already put forth and that initial guidance from M-24-10 as a big part of what the executive order outlined for this new guidance to achieve,” she added.
Executives across Federal agencies are going to be busy in the next couple of months meeting the deadlines of both of these memos. The first deadline of Nov. 1 comes from OMB’s most recent memo, M-24-18, and directs Federal agencies to identify any contracts associated with agency use of rights-impacting or safety-impacting AI.
However, Hubner noted that while Nov. 1 “is not too far away,” the requirement should not come as a shock to agencies.
“To be very specific, Section 50 of M-24-10 was all about the way to think about procurement from an AI risk management perspective,” Hubner explained. “So, agencies have known that contracts that are written in a way that preclude compliance with M-24-10, for any reason, would need to be renegotiated or modified in some ways to make sure that they had what they needed to then comply with the documentation requirements of M-24-10.”
“So, that is not a new deadline in the sense of what is being accomplished, in terms of what is happening at the agencies. What is perhaps a little bit more different is just that we’re putting it more in writing,” she said, adding, “More as just a benchmark for agencies to have to look at and be like, ‘Alright, this is how I want to set myself up for success to then hit that Dec. 1 deadline.’”
The Dec. 1 deadline comes from OMB’s March memo, M-24-10, which tasks Federal agencies with ensuring all of their safety-impacting and rights-impacting AI use cases comply with M-24-10’s risk management requirements.
“It is going to be a very busy next few months, but OMB is continuing to work really closely with agencies to ensure that they have the tools they need, the information they need, and the access to us that they need to really help them accomplish this really meaningful work,” Hubner said. “I think it’s easy to get caught up in the policy aspect of it and the implementation aspects, but at the end of the day, it’s the American public’s rights and safety that are being impacted when AI is used in these use cases.”
“There’s actually a lot of passion, I think, throughout the interagency to really bring this work together. So, it’s going to be busy, but it will not be without, I think, significant reward,” she said. “It’s going to be a really strong win for just the government writ large to get this accomplished.”
