The Biden administration’s recent artificial intelligence (AI) executive order (EO) is rapidly accelerating AI use and employee training among Federal agencies, a top Cybersecurity and Infrastructure Security Agency (CISA) official said on Nov. 15.
The order has served as “a tremendous accelerator to making sure that every single agency across the Federal government is taking AI seriously and investing in a workforce that is ready to tackle these issues,” Lauren Boas Hayes, senior advisor for technology and innovation at CISA, said today at the Red Hat Government Symposium: Intelligence Unleashed.
Although the flurry of recent news about AI has created somewhat of a “hype cycle,” Boas Hayes said, the technology provides “incredible transformative opportunities” for all agencies, especially those in the national security and defense realms.
At CISA, she added, the EO has helped spur efforts to work with industry and international partners on development of AI tools and also to begin the type of “red and blue team” exercises the agency has been undertaking to help identify malicious cybersecurity actors.
“We are focused on red teaming,” Boas Hayes said. “And now the question is, what is AI red teaming? And how do we make sure we’re part of the development of that industry?”
Her remarks came a day after CISA released a Roadmap for Artificial Intelligence policy document that aims to promote responsible usage of AI in the agency’s mission to protect Federal civilian agencies and critical infrastructure sectors, while also assisting government and private sector organizations in securing the AI-enabled software they use.
The roadmap was one of a number of agency follow ups to the administration’s Oct. 30 AI executive order, which focused on seizing the promise and managing the risk of the emerging technology. The EO established new standards for AI in eight categories, including safety and security; privacy; equity and civil rights.
Boas Hayes spoke at a Red Hat symposium modernization panel about how to “Unleash the Promise of AI With End-to-End Data Science.”
She was joined at the session by Damian Kostiuk, deputy chief data officer at U.S. Citizenship and Immigration Services. (USCIS).
Kostiuk emphasized that for all of the anticipation about integrating AI technologies in government, the best way to develop trustworthy AI is to start with “working through the kinks of the easy stuff.”
By that, he said he meant sharing data across existing data silos and figuring out how AI can mesh with the legacy technology still in use in homeland security and other agencies.
“There are some different directions you can go with the trustworthy route, and you start with that as an orientation,” Kostiuk said. “Do you start with openness and transparency as an orientation? Are you trying to hit a lot of these different elements? Honestly, you kind of have to hit all at once.”