Miranda Bogen

My thoughts on the Senate AI Working Group's new AI Framework: “We have to wait and see what bills and roles are proposed based off of a lot of this but a lot of it is indeed promising. There's a strong emphasis on R&D, there's a strong emphasis on regulatory clarity, for example, and making sure that our existing laws are clear so that we can ensure easy, confident adoption of these technologies. There was clearly a thoughtfulness that went into this,” Mittelsteadt said. I'd estimate around 75% of the package is headed in the right direction: emphasizing engineering-based solutions, clearing up laws, and making targeted spending bets on talent and other discrete priorities. That said, elements such as AI export controls solutions in search of a problem and a recipe for policy failure. Too much is on the line to overzealously regulate. While action is needed on many fronts, with regard to export controls: “They've been there a broken record since at least 2016 about social media regulation, if not earlier. And when you cry wolf for going on 10 years about technology regulation, people stopped believing you,” said Matt Mittelsteadt, a research fellow at George Mason University’s Mercatus Center. “That's exactly where Congress is, and I think the most likely thing that we can expect is nothing."

4

A nice write up about my new case study on Federal AI use rules and how they are suppressing the ability of agencies to manage AI risks/benefits in their contexts. Specifically, I discuss the federal governments deep translation capacity problems which are yielding life-impacting service , intelligence , and emergency response failures. Machine translation comes with error but could also fill gaps when the alternative is no translation at all. Unfortunately, one-size-fits all federal AI minimum "risk management" standards inhibit the ability to weigh risks and benefits in the context of a program/agency. The result: any benefits are lost and the many, many negatives of exsisting capacity failures locked in. https://lnkd.in/e-_s5H3q

6

1 Comment

 My latest in Federal News Network discussing key priorities for the newly created Artificial Intelligence Safety and Security Board. 1. I might sound like a broken record - but the first step will be to prioritize targets. "Critical infrastructure" has massive scope and members should concentrate core systems including the grid, water, energy, transportation, and healthcare. Focusing on these essential areas will allow for deep consideration and effective action. 2. Prioritize preparation for cyber. Focus on analyzing existing regulations, identifying gaps, opportunities for harmonization, and ensuring government agencies are adequately resourced to address potential AI-related threats. While such threats have yet to emerge in force, policy should focus less on trying to avert specific threats and instead on preparing for unknown. We don't know what may come down the line but building the capacity for nimble analysis and well-resourced action will be key to future success. 3. Prioritize Diffusion. Recognize AI's potential to improve reliability and efficiency in critical infrastructure by studying bottlenecks, regulations, and research needs that hold back success. AI applications such as predicting wind power supply, detecting downed powerlines, and augmenting cyber defenses could significantly benefit these systems. By concentrating on these priorities and developing clear strategies, the AI Safety and Security Board can ensure much needed trust and maximum AI success in our nation's most essential systems, leading to a more prosperous future. https://lnkd.in/eRrZKSew

5

My thoughts quoted in Inside AI Policy on the GOP's stated intent to repeal the Biden AI Executive order and instead, in their words, "support AI Development rooted in Free Speech and Human Flourishing." “Matt Mittelsteadt, also of Mercatus, said of EO repeal " The question is: what replaces it? The previous Trump administration pursued a light touch approach, but the AI context of 2020 is a far cry from 2024. While it's unclear what AI development promoting ‘human flourishing’ might mean to the Trump camp, known Trump policy priorities such as expanded tariffs, expanded border controls and a general ‘America first’ attitude could put the American AI industry at a competitive disadvantage and slow AI innovation and investment.” https://lnkd.in/dEG3C29X

6

“This report and accompanying website https://aspi.org[dot]au/techdiplomacy outlines the current state of play in global AI governance, looks at the role of technical standards, explains how agreements on technical standards are negotiated and created, and outlines who are the biggest ‘movers and shakers’.” https://lnkd.in/gA4nJYcC

1

Round II with Charlie Mitchell and Inside AI Policy! This time, a discussion on how #artificialintelligence policy in the incoming Administration may be impacted by the concept of a new possible AI Czar. Happy to have shared a few thoughts with Charlie, here are a few key items, and the article is below! 🔸 "The oversight and direction on AI policies and programs across all federal agencies is a major effort,” commented Norma Krayem, vice president and chair of Van Scoyoc Associates’ Cybersecurity, Privacy & Digital Innovation Practice Group and a former senior federal official. 🔸 “To achieve this, it is important for both executive branch agencies and public/private stakeholders to know who is in charge to ensure cohesion and to understand what changes may occur in policy as well,” Krayem told Inside AI Policy. “Unlike other technology, once AI is integrated into programs it is not easily undone.” 🔸 Krayem said, “As the new incoming administration reviews these issues, if they choose to designate an ‘AI Czar,’ that person needs to work hand-in-hand with the Office of Management and Budget, Office of Science and Technology Policy, National Security Council, National Economic Council and others to ensure there is a cohesive approach on implementation of all AI policies.” 🔸 Krayem pointed to Congress’ interest in AI issues, including through the bipartisan Senate working group and House task force, and urged lawmakers to continue those efforts. #AIpolicy #criticalinfrastructure #Congress #globalpolicies Van Scoyoc Associates

4

A reminder that the Urban Institute and the Microsoft Justice Reform Initiative will host an informational session on October 16 about the latest request for proposals in the Catalyst Grant Program.  Register here - https://lnkd.in/eGV7s9uq      We will post recordings of the workshops on the grant website after the sessions.        You can visit the grant application website at https://lnkd.in/ex499AhD to download the full Request for Proposals (RFP). The application is responses to 10 narrative questions about the proposed activities and organization, totaling an estimated 5 pages of text.   Proposals are due on November 22.       Background: The Catalyst Grant Program provides #funding and technical assistance to #nonprofit organizations in the United States for projects that use data and technology to advance racial #equity and reform in the adult criminal legal system. We are seeking local projects that support policies, programs, and practices that prevent unnecessary system involvement and decrease racial and ethnic disparities in the front end of the criminal legal system. Lead organizations must be 501(c)(3) nonprofits (universities and government agencies can be partners but not leads).  You can learn about grantees from earlier cohorts on the Urban Institute website at https://lnkd.in/eZ-cin4H   

27

1 Comment

Interested in the impact of #ArtificialIntelligence on US homeland security policy and critical infrastructure protection? Check out this fascinating blog from my fellow National Security Institute (NSI) Fellow Jeffrey W.. It’s definitely a #mustread! #artificialintelligence #AI #DHS #homelandsecurity #criticalinfrastructure #nationalsecurity #thescif

6

1 Comment

Please read, because the conclusions are counter intuitive.

11

The 2024 Election Results & #AI Policy: Here Comes the Mother of All State Regulatory Patchworks my quick thoughts on what the election results mean for AI policy going forward at the state level. More thoughts to come about what happens at the federal level, which is far more uncertain. https://lnkd.in/dAAWhnyX #artificialintelligence #machinelearning #techpolicy #regulation #innovation #states #federalism #preemption

18

1 Comment

I'm keeping an open mind about the potential for AI to impact CBRN threats and generally encourage efforts to measure these issues in order to ground the conversation and minimize often-fantastical dooms-day scenarios. That said, evidence that AI is meaningfully changing current CBRN threat levels is lacking and probably overstated. Beyond this report, DHS must consider how to best prioritize its limited AI bandwidth. With a fixed budget, any excess money spent on mitigating or analyzing these hypotheticals nessesarily comes at the expense of other, evidence-backed priorities. Rather than over-invest in AI-CBRN issues, DHS should spend on the following: 1. Cybersecurity and other evidence-based challenges: while early days, there is indeed evidence that AI has changed the cyber game. In the past year spear phishing increase 1,200%, machine translation improved the quality of social engineering attacks, and state backed actors have used LLMs to generate operational malware. While not quite a crisis, this is a meaningful change that could portend a real challenge if capabilities continue to improve. Rather than chase CBRN 'what ifs,' resources should focus on this very real, evidence-backed challenge. 2. Preparing for the unknown: Its still early days, the tech is changing rapidly, and its uncertain what challenges will be the core AI issues of the coming years. Rather than overinvest resources on CBRN threats, DHS should invest in the agility and capacity needed to quickly identify and react to a range of unexpected, emergent challenges. The first step is state capcaity, DHS should focus on upskilling and hiring to build a work force that not only understands AI, but AI's intersection in a broad diversity of domains. Second is regulatory flexibility, DHS should actively analyze its rules to ensure it can act with minimal bureaucratic delay if we see a black swan AI development. Third is vigilance, DHS should foster the infrastructure, such as Information Sharing and Analysis Centers, needed to quickly identify unexpected changes in the threat environment. It is indeed worthwhile for the Countering Weapons of Mass Destruction office to analyze how AI could impact their work. Without evidence, however, this shouldn't become a preoccupation. As the DHS proceeds it should focus limited bandwidth on proven challenges and for all else, build the threat-agnostic capacity needed to act with agility.

7

Helpful tool to understand Americans' use of federal websites. At the top of the page, you can specify a specific department, e.g. HHS: https://lnkd.in/gBKvuKv8

2

Comments for Tech Policy Press on this week's report from the Task Force on Artificial Intelligence led by co-chairs Jay Obernolte (R-CA) and Ted Lieu (D-CA) laying out recommendations on AI: This week’s bi-partisan Congressional Report on AI highlights a critical challenge: while AI systems are rapidly becoming a backbone for decision-making across sectors, we cannot yet be sure they are fair. While it is frequently assumed that AI fairness is about preventing historical forms of discrimination, the truth is more complex: AI systems can develop unexpected, and even arbitrary, preferences. As we expand AI deployment across housing, employment, education, and criminal justice, assuring basic fairness is an essential safeguard. Without it, we will undoubtedly have consequential but unfair outcomes that are both undetected and uncorrected. The Report also rightly flags the profound relationship between uses of AI and civil liberties. In the current moment, we must recognize that our widespread lack of data privacy and foundational data minimization poses the real “x-risk” for this technology, given the threat of manipulation of public opinion, including micro-targeting of highly tailored disinformation, the growing use of facial and bio-metric recognition tools, and the consequences these all may have when powered at scale for a fair, free, and democratic society. https://lnkd.in/exbXZznr

6

Last week, we filed supplemental testimony with the United States Commission on Civil Rights on the use of Artificial Intelligence by the Department of Homeland Security (DHS)-- we close them with a what we hope is a useful set of questions for DHS: 1) What stakeholder engagement is planned around specific use cases? 2) What empirical testing has been conducted with real-world users (e.g., has TSA measured whether or not airline passengers fully understand their ability to withhold consent at checkpoints to FRT uses? What evidence suggests that they do or do not feel at liberty to make a choice?) 3) What changes in biometric surveillance policy or other steps are DHS considering, if any, in light of the OMB EO? 4) What programs at DHS will likely receive waivers in December and what is the timeline for aligning those uses with the requirements of the EO? 5) What are the implications for FRT of Secretary Mayorkas’s new AI initiative? 6) What evidence of consent for data collection does DHS regarding its well-publicized collections of DMV and utility data that include US citizens? 7) Why does DHS feel a need to hold onto non-citizen data for an astonishing 75 years? 8) What training has been completed at this stage in terms of the timeline and what has yet to be done? 9) Does the planned training adequately capture the human factors and limitations of the tech use cases? Will the training be shared with the Commission? Has it been independently evaluated? 10) What are safety evaluations that have been conducted by DHS regarding risks to migrants and new, more hazardous corridors that may be created in light of the plan to expand DHS surveillance technologies and towers at the border? https://lnkd.in/eM3K7vm7

3

Great to talk to Charlie Mitchell, Inside AI Policy about possible changes to #AI policy across the U.S. Govt. in the incoming Administration. This piece focused how OMB could address these issues in light of the recent news that former OMB Director Vought would return as the new OMB Director. A few points: 🔹 “In November 2020, then Director Vought issued OMB memo M-21-06 which laid out key policy, legal and regulatory approaches for AI usage in the U.S. government,” “It focused on the need to ensure that economic and national security issues, alongside privacy and civil liberties, human rights and the rule of law were considered by agencies when using AI, as well as with the broader implementation of technology and innovation.” 🔹 “The same themes around AI usage today remain critical, as were in that memo: the importance of safety and security, the need for public trust in AI, public participation, risk assessment and management, scientific integrity and information quality are all still crucial to manage. In fact, considering the leaps and bounds in AI development in the last four years, those core tenets are even more critical to maintain and support as we go forward.” #technology #governmentcontractors #responsibleAI #safesecureAI #cybersecurity #criticalinfrastructure Van Scoyoc Associates #artificialintelligence

8

1 Comment

The senate AI working group released their long awaited framework for AI legislation. Matching AI’s broad-based impact, this is an incredibly broad package. Overall this is decent, however the many positives are paired with a few concerning lows. Here's some highlight on what's inside: On the upside there is a clear emphasis on promoting innovation and diffusion including: 1. Talent development through possible upskilling incentives, federal talent development programs and fellowships, and security clearance fast tracking. 2. State capacity building through funding for IT modernization, public private partnerships to tackle key issues, and possible new voluntary survey to study AI economic impact and key issues. 3. Regulatory clarity though a regulatory gap analysis in the financial sector and a report from the comptroller general on “any significant federal statutes and regulations that affect the innovation of artificial intelligence systems, including the ability of companies of all sizes to compete in artificial intelligence.” On the downside we find: 1. Export controls: supporting the recently introduced ENFORCE Act, the Senate is considering AI export control authorities. It’s unclear what problem controls could solve, but almost certainly would create needless challenges for software and AI companies alike. 2. Possible pre-deployment certification: while non-committal the framework also opens the door to requirements for possible govenrment review of certain models. What this might look like, and how serious the senators are is unclear. That said, such testing is largely to be avoided. 3. Restricted Data: A dose of AI apocalysm with possible classification rules modeled on the Atomic Energy Act’s restricted data regime. This worryingly conflates AI and atomic materials. Restricted data classifies information upon existence. Absolutely discouraged. On top of these we also find a smattering of largely disconnected issues including online child safety, outer space security, and addressing the opioid crisis. As a general rule, any legislation should focus on the task at hand: AI. Lets avoid making any AI leg a congressional catch-all. All of this considered, I’m optimistic. The vast majority is productive or benign and provisions such as the CREATE AI Act’s research support are where we find the most support. That said, senators must avoid harmful restrictions and controls without empirical evidence of real problems. https://lnkd.in/eN6vqUGJ

3

To all those who saw or followed our post about the new U.S. Congress dataset on Hugging Face for policy areas and subjects... At the suggestion of Marci Harris of POPVOX Foundation, our team at DreamProIT now created a new dataset that includes Congressional committees: https://lnkd.in/gzACiADG With this data set, it should be possible to suggest committees to refer any bill to. As with subjects and policy areas, automating recommendations would not replace the judgment of subject matter experts, but could provide a baseline to show what committee would be assigned based on prior precedents of the House and Senate Parliamentarians. #data #uscongress #library

34

6 Comments

Federal CDOs are making tremendous progress in achieving their missions & building data maturity across their organizations. But they need more resources & clarity around their responsibilities - particularly in light of a growing focus on AI adoption in government. Excited to be a part of Data Foundation 's fifth survey of CDOs in collaboration with Deloitte Government & Public Services, check out our new report on the results out now!

6

This webinar is today!

4