Alejandro Mayorkas held what’s been referred to as “the hardest job in Washington” throughout a very turbulent time in U.S. politics. As Biden’s Secretary of the Division of Homeland Safety, he endured an impeachment by House Republicans, a fentanyl crisis, a string of natural disasters, and plenty of terror attacks on house soil. However he additionally ran the sprawling division throughout a serious expertise shift with generative AI, and was charged with leveraging the expertise to make DHS extra environment friendly and responsive. Politically, Mayorkas could also be remembered as an immigration lightning rod (Mayorkas himself immigrated to the U.S. from Cuba as a baby), however throughout the DHS he’ll even be often known as the man who combined AI with paperwork.
In an exit interview with Quick Firm throughout certainly one of his closing days at DHS, Mayorkas provides his spotlight reel of the division’s AI work, which included working with tech sector companions together with Sam Altman and Jensen Huang, and recruiting AI engineers in Silicon Valley. The interview has been edited for size and readability.
The generative AI growth got here throughout your time period as secretary. How properly did DHS reply to this large expertise shift?
I’ll say fairly candidly we responded fantastically, and I say that giving credit score to extraordinary folks right here within the division. We moved swiftly and expansively, and we have now led the federal authorities in embracing the potential of AI in all regards: to harness its potential to advance our mission, to grasp the potential for its malevolent use and start to take steps to guard the homeland and most notably crucial infrastructure towards that problem, to construct a workforce that’s able to each harnessing and securing AI, constructing partnerships with the personal sector, and promulgating insurance policies to control using AI into the longer term.
Having a major position in several regards which are codified in presidential govt orders, we have now completed so very a lot. We moved in a short time in growing our AI highway map early in 2024. We issued insurance policies and procedures governing our use, we assembled the AI Security and Safety Board that I chair with luminaries from all through the AI ecosystem, and we created the place of chief AI officer [currently held by Eric Hysen]. We employed 49 skilled technologists to kind our AI Corps. I might go on and on.
Are there any moments that you just bear in mind being notably joyful, like seeing this expertise utilized and making a distinction on the bottom?
I can establish plenty of moments, however when it comes to seeing AI in motion, it’s outstanding to see us having educated a machine to simulate a refugee applicant with the intention to present coaching for our refugee officers. The coaching yielded not solely a machine that might reply to the officer’s questions substantively when it comes to nation situations and what the machine applicant lived by means of, but in addition educated the machine in how a refugee applicant would reply the questions given the trauma endured. For instance, coaching the machine to be reticent in offering sure particulars—it’s extraordinary. I cite that as a result of I’m a political refugee myself. I got here to this nation as a one 12 months previous, however actually my mother and father shared with me their refugee experiences and so it was fairly resonant.
I perceive that there are actually 158 AI functions in use throughout the division, together with 29 programs which are straight impacting immigration processing and doc processing.
We’re taking a look at our work throughout the spectrum all through the division and assessing the place we might deploy AI to advance that work. There’s no stone left unturned within the exploration of the potential for AI in our work.
It looks like there’s lots of friction within the processing of individuals coming to this nation. It requires gathering lots of data and documentation. Is that one thing AI would possibly remedy?
It ought to be used to drive transformative effectivity. I’ll inform you once I entered the Immigration Company, U.S. Citizenship and Immigration Companies in 2009, it was purely paper based mostly. I imply, there was a small quantity of on-line functionality, however typically aerospace. And that is the twenty first century and so we have now modernized our processes significantly. Now we have now AI which might take that to a really completely different degree, and while you communicate of friction, I’m pondering of bureaucratic friction, if you’ll, and the elimination of that. The political friction—that will be magical if AI might handle that.
Your chief AI officer, Eric Hysen, and his crew produced a Generative AI Public Sector Playbook that accommodates learnings from all the things DHS has constructed with AI. How do you hope the doc may be used sooner or later?
Now we have printed plenty of seminal paperwork as a result of our focus has been on institutionalizing this dedication to the potential of expertise writ massive, and clearly AI could be probably the most transformative expertise in a very long time. We thought it each prudent and necessary to publish paperwork that may assist institutionalize this dedication and hopefully safe the longevity of this dedication.
This 12 months you probably did lots of work to deliver collectively the Synthetic Intelligence Security and Safety Board, which incorporates a powerful lineup of names that we’re accustomed to in AI circles. Is that board going to outlive by means of the subsequent administration?
That could be a query that ought to be directed to the incoming administration. It’s actually my hope that it continues. Now we have, in collaboration with that board, produced a groundbreaking framework for the protected and safe deployment of AI in crucial infrastructure, and that board is apolitical. As you famous, it’s constituted by luminaries within the tech business, in crucial infrastructure, in civil society, and in public workplace.
And the board produced the Framework for the Protected and Safe Deployment of AI in Vital Infrastructure, launched in November.
It was a big funding of thought and time and power on the a part of this division, me, and plenty of folks right here, and the board members and their respective groups.
Turning focus onto the recruiting you probably did to employees the AI Corps, I used to be particularly to listen to that you just personally went to the West Coast to interview candidates. Via that have, what did you study how Silicon Valley folks take into consideration authorities service?
What I did in my recruiting efforts in Silicon Valley and elsewhere within the nation was share with them what we do, the impression on folks’s lives in a myriad of how, and invited them to be part of it. To not consider authorities from a 50,000-foot perch of rhetoric, however slightly I defined to them what we do this impacts folks’s lives, makes folks’s lives higher, and what an unimaginable privilege it might be for them, and the way fulfilling and galvanizing it might be for them to be part of it. We had been recipients of a wave of functions, 1000’s if I’m right.
That applies to the way in which you discuss AI to DHS folks internally too, I think about.
Sure, and I have to say, our workforce could be very excited concerning the potential of AI. We wish to be clear that we’re leaning ahead considerably with applicable guardrails to make sure that our use of AI is accountable, that it’s respectful of our commitment to civil rights and civil liberties and privateness pursuits.
I do know accountable AI has been a key concern however, you probably did obtain a letter from plenty of immigration civil rights teams in September saying that DHS is quick monitoring these AI functions with out sufficient regard for transparency and civil rights. Did you meet with these people and what was the consequence?
Sure, we met with them and despatched a response. The letter has errors of truth and in addition accommodates a misunderstanding of what we do and don’t do, and so the letter we set forth in response, I believe will lay it out fairly clearly. There are lots of errors and misunderstandings in that letter.
They had been notably involved about ICE utilizing biometrics knowledge with AI. Are you able to touch upon that?
I’ll seek advice from the blog post and the responsive data as a result of we handled every.
I do know you’ve been talking with each [DHS secretary nominee] Governor Kristi Noem and the Trump transition crew. Was AI part of these discussions?
It’s not for me to talk of the conversations I’ve had with Governor Noem or that our crew has had with the transition crew. I’ll say that I’ve spoken with Governor Noem plenty of occasions and have been each impressed and appreciative of these conversations.
Among the DHS AI initiatives grew from the White Home’s executive order on AI. Some Republicans in Congress have already expressed an curiosity in reversing the EO. Might that derail a number of the division’s efforts or endanger DHS funding for AI work?
I don’t know what actions will and won’t be vis-à-vis the manager orders. However I’ll inform you that I’m optimistic that the incoming administration will proceed to make use of expertise to advance authorities missions. I’m fairly optimistic about that.
