What You Need to Know About the New AI Executive Order

Level Set: Unwrapping EO 14110

The Administration recently released the Executive Order (EO) on Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence order – or Executive Order 14110. Its purpose is to lay out principles and guidelines for AI at a more developed level than have been available to date for the federal government. The EO is thorough and a lot to take in, so The Clearing’s data and technology experts developed a primer to help federal leaders get up to speed so they can begin thinking about how the EO will impact their agencies.

Setting the Tone

As the first EO focused on AI, it does a solid job balancing a tone of possibility with the potential risks of unchecked AI. It recognizes the significance of this new technology, the fact that it is now widely accessible, and its power and potential – for both good and harm.

Nuts, Bolts, and Concerns

Executive Order 14110 covers a lot of ground. It includes the nuts and bolts of federal AI use and adoption; however, potential concerns – and their possible mitigations – also come through prominently. Here are several points our team found noteworthy.

1. The order includes a call for a concerted government effort towards standardization, including in the development of tools and methods for ensuring security and safety, and identifies the National Institute of Standards and Technology to lead this effort.
2. To better assess security concerns, the EO proposes requiring developers of AI models and systems to provide the government with the results of their red teaming exercises to identify vulnerabilities and dangers within their tools. How that data is shared will likely be a major point of discussion between the public and private sectors.
3. While national security is a primary focus of the EO, social and workplace issues are also covered with a particular focus on protecting privacy. To counter this risk, the EO includes a directive for funding and developing protective mechanisms against accessing and exposing the private information of the American people.
4. Our team was happy to see that the order recognizes the need to ensure AI isn’t used for discriminatory purposes or encroach on civil liberties. For example, the order highlights how AI could be used by benefits programs, landlords, or other agents in ways that could exacerbate discrimination. In addition, there is a call for creating clear guidelines around how AI is deployed within the criminal justice system so that it doesn’t tilt that system in a damaging way toward either prosecution or defense. These provisions show just how wide-ranging the potential for AI is and that no federal agency – regardless of mission – will be immune from it.
5. Finally, the EO weighs in on worker impact, including funding programs for retraining employees and mitigating the impact that AI will have on the labor market. This includes a call for studying these impacts, as well as coordinating with our allies and trade partners on the approach to dealing with AI in the labor market.

The Effect on Federal Agencies

The transformative impact of Artificial Intelligence on federal agencies is undeniable. Notably, the Executive Order (EO) is poised to exert a considerable influence, particularly in shaping security and privacy measures during the EO’s development and implementation.

Here is more on how I see these unfolding.

• Hiring needs. There is already an emphasis on hiring a significant number of AI professionals. While agencies charged with developing the tools and policies surrounding AI-focused security threats will feel the pressure to hire first, every agency will eventually need to augment its workforce with AI professionals to ensure these policies are implemented successfully.
• Real-time adoption. Even with the best of planning – and this order is a start – employment of AI in the federal space will be a challenge because this is happening in real-time. There is no line in the sand; people and agencies are already using AI and adapting along with it. That means implementing the standards called for in the EO is critical, but makes doing so even more challenging.
• Data management. Many of the components of this order are premised on gathering data about AI and its potential impacts in all of these different areas. I suspect there will be a lot of data work around this order. Given existing challenges in data management, use, and governance, agencies will be forced to appropriately resource their data strategies while adapting to the real-time challenges posed by AI.

What’s Next

As is true for most Executive Orders, there is more to come. Because EO 14110 explicitly calls for the development of additional understanding and specific instructions for how to manage AI, combined with the speed of AI development, federal leaders need to be acutely aware of how fast this is going to be coming. This will likely lead to unique challenges for federal tech leaders such as CIOs, who will have to simultaneously be flexible and navigate a nascent regulatory environment.

It’s a lot to think about and plan for. If you find yourself needing more insight on Executive Order 14110, AI, or data management, The Clearing’s data and technology teams are here to help. Reach out anytime – we would love to help you get ahead of the game.