Senators introduce bill on federal agency procurement and use of AI — AI: Washington Report | Mintz – An Antitrust Perspective

Applications of AI


[co-author: Matthew Tikhonovsky]

  1. Senators Gary Peters (D-Mich.) and Thom Tillis (R-North Carolina) introduced the PREPARED for AI Act.
  2. The bill aims to ensure federal agencies can safely and effectively deploy AI technologies by establishing a comprehensive risk mitigation framework for the procurement and use of AI.
  3. Given the bill's bipartisan nature, and reports that senators plan to make amendments to the bill in committee this summer, it is possible the bill could move forward this summer.

On June 11, 2024, Senators Gary Peters (D-MI) and Thom Tillis (R-NC) introduced the “Preparing for Enterprise-Wide Deployment of AI Act” (the “PREPARED for AI Act”).

The PREPARED for AI Act aims to ensure federal agencies are able to responsibly and effectively adopt the latest AI technologies. The bill establishes a risk mitigation framework for federal agency procurement and adoption of AI, and creates governance structures and programs to ensure federal agencies can benefit from advances in AI.

The Act builds on the privacy and civil rights protections established by the Advancing America AI Act of 2022 and the AI ​​procurement and use regulations finalized by the Office of Management and Budget (OMB) in March 2024, while also creating additional safeguards and programs. Following last week's introduction of the bill, below we summarize recent efforts to regulate and advance the use of AI by federal agencies and analyze a number of provisions in the PREPARED for AI Act.

Safe and effective use of AI

As federal agencies increasingly use AI, several lawmakers have pushed to ensure they can leverage the latest AI technologies while avoiding AI-related risks, including privacy and civil rights concerns. The PREPARED for AI Act builds on three existing developments in federal AI policy:

The Advancing America's AI Act, included in the Fiscal Year 2023 National Defense Authorization Act, aims to balance innovation and security in federal agencies' adoption of AI by encouraging agencies to accelerate the adoption of new AI technologies through pilot programs and requiring agencies to identify and mitigate risks associated with these technologies.

As we wrote in October 2023, President Biden's comprehensive Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence was focused on promoting the responsible and effective use of AI. Among its provisions on AI innovation and safety, the Order tasked the Office of Management and Budget (OMB) with issuing guidance on the use and procurement of AI by federal agencies.

Pursuant to this Executive Order, OMB issued final guidance to federal agencies on the procurement and use of AI in March 2024. As mentioned above, OMB guidance strengthened agency AI governance, instructing agencies to inventory and categorize AI use cases and conduct appropriate due diligence regarding AI procurement. These OMB policies may be subject to change or repeal by a future administration that takes a different approach to AI regulation.

AI Countermeasures Bill

The PREPARED for AI Act aims to establish in law a risk mitigation framework for the safe and effective procurement and use of AI by federal agencies. The bill's requirements fall into four broad categories, which are described below.

  1. Strengthening AI governance
    1. Establishment of a Council of Chief Artificial Intelligence Officers (CAIO): The CAIO Council will be made up of each agency's CAIO and will be jointly tasked with coordinating agency AI development and use, sharing best practices for AI innovation, managing risks associated with AI procurement and use, and reporting incidents related to AI procured or used by the agency.
    2. Establishment of the Agency's Artificial Intelligence Governance Committee: Each agency's AI governance committee would be convened by the agency's chief information officer (CAIO) and would “coordinate and govern artificial intelligence matters across the agency.”
  2. Using AI to reduce risk
    1. Building an AI risk classification system: Agencies will be required to develop a risk classification system for AI use cases, with at least four classification levels (unacceptable, high risk, medium risk, and low risk). The high risk classification applies to AI systems that affect decisions regarding civil rights, equal opportunity, “ability to access or apply for significant government resources or services,” “individual or community well-being,” the environment, and other categories.
    2. Protecting and monitoring high-risk use cases: Agencies that adopt AI technologies classified as high risk must closely monitor the ongoing use of these technologies, and these agencies must create reporting mechanisms for adverse incidents related to high-risk use cases and requirements around sourcing, managing, and protecting data.
    3. Prohibition of certain artificial intelligence use cases: Government agencies will be prohibited from using AI facial recognition mapping to assign emotion, assess trustworthiness, infer race or protected beliefs, etc. Use cases of AI that are classified as unacceptable will also be prohibited.
  3. Mitigating Risk in Federal AI Procurement
    1. Risk assessment and pre-implementation requirements for using AIEach agency's CAIO would be required to develop a risk assessment process to assess the risks of AI systems before they are purchased and used by the agency, and would be required that deployers and developers of AI used by federal agencies submit information to the CAIO regarding the AI ​​architecture, sources of data, data management, privacy practices, and other relevant information.
    2. AI contract requirements: Government contracts for AI include mandatory requirements for safety, security, reliability, data procurement and ownership, AI evaluation and testing, and documentation.
    3. Annual report on obstacles in AI testing: The CAIO would be required to submit an annual report to the CAIO Council on “impediments to artificial intelligence testing and evaluation” and provide recommendations to address those impediments, and would convene a working group to “develop tools and guidance to assist agencies in addressing the impediments identified in the report.”
  4. Benefit from the latest AI advancements
    1. Creating a multi-phase commercial technical test program: The program creates a pipeline for government agencies to solicit proposals for AI contracts and then test and evaluate the proposals through a multi-stage competitive review process.
    2. Establishment of a pilot program for R&D projects: The pilot program will allow government agencies to conduct AI research, development and prototype projects through a competitive contracting process.

Conclusion: The Future of AI Readiness Act

After the bill was announced, many relevant stakeholders, including the Center for Democracy and Technology, the Transparency Coalition, AI Procurement, and the Institute of Electrical and Electronics Engineers, have come out in support of the bill. Senator Peters, who serves as the Chairman of the Senate Homeland Security and Governmental Affairs Committee, reportedly plans to hold a committee markup of the bill over the summer.

However, while the bipartisan nature of this bill is likely to gain traction later this summer, it is by no means certain that it will pass, especially before the end of this year's Congress. There is no indication that the House of Representatives will take up this or a similar AI bill. As we've discussed in previous newsletters, there are many obstacles to federal legislative action on AI in the near future. The Senate has been holding AI Insights Forums for months but has yet to produce any legislative proposals, and as we discussed earlier this month, specific and far-reaching AI legislation remains a topic of debate.

Despite these concerns, interested parties should still closely monitor the progress of the PREPARED for AI bill, which, like Senator Peter’s Advance America’s AI Act of 2022, could be incorporated as part of a broader package such as the National Defense Authorization Act or a comprehensive AI bill, or it could serve as a starting point for these themes in 2025.

We will continue to monitor developments on this bill, as well as other AI-related developments in Washington.

[View source.]



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *