Letter to the White House on Forthcoming AI Executive Order
Dear President Biden and Vice President Harris:
We, the undersigned civil, technology, and human rights organizations, write in response to the Biden-Harris administration’s July announcement regarding the development of an Executive Order on Artificial Intelligence (AI EO).[1] We urge the administration to focus this AI EO on protecting the American public from the current and potential harms of this technology—including threats to people’s rights, civil liberties, opportunities, jobs, economic well-being, and access to critical resources and services.
Thanks to this administration’s foresight, the White House’s own Blueprint for an AI Bill of Rights (AI Bill of Rights)[2] provides a roadmap for how to implement the responsible use of AI across the federal government. The AI EO takes on even greater importance as we anticipate the Office of Management and Budget’s (OMB) draft guidance on AI use by federal agencies.[3] Protecting civil rights and civil liberties, ensuring privacy, and advancing equity should be at the center of the AI EO and OMB guidance memo.
We appreciate the initial actions taken by the Biden-Harris administration to address the challenges and opportunities of AI. Most notable is the year-long process of extensive stakeholder engagement with industry, civil society, academia, and government that led to the development of the AI Bill of Rights.[4] The AI Bill of Rights articulates essential principles that can adapt to the fast-paced development of new technology and lays out detailed practices for government and industry that, if implemented, will ensure Americans are protected from the harms of automated systems and equitably benefit from their possibilities. These are necessary measures that will enable the administration to achieve its stated goals of managing AI risks and protecting individuals’ rights and safety.
The AI Bill of Rights’ five principles—Safe and Effective Systems, Algorithmic Discrimination Protections, Data Privacy, Notice and Explanation, and Human Alternatives, Consideration, and Fallback—and accompanying technical, regulatory, and policy practices are essential guardrails that will secure the rights and ensure the safety of the American public and those workers impacted by these technologies.
A new AI EO should make the AI Bill of Rights binding U.S. government policy, order the U.S. government to implement the best practices outlined in the AI Bill of Rights, and ensure that federal agencies and contractors cannot deploy or must stop using automated systems that do not abide by these principles and practices.[5] Federal agencies funding, acquiring, or using an AI system have a responsibility to ensure that the system works and is fit for purpose. AI systems that are biased are broken and not fit for purpose, and the decisions made using those systems cannot be trusted.
Simply put, the federal government should not use an AI system unless it is shown to be effective, safe, and nondiscriminatory. AI should work, and work for everyone. Toward that end, we offer the following recommendations:
We urge the President to make the White House AI Bill of Rights binding on the use of AI tools by all federal agencies, contractors, and grantees in the forthcoming AI Executive Order.
The new AI EO should direct the executive branch to immediately implement the AI Bill of Rights for federal agencies, contractors, and grantees.[6] The federal government is the largest employer in the country, one of the largest purchasers of technology, and has enormous ability to shape the emerging AI policy and business landscape. Committing the executive branch to the AI Bill of Rights and its five core principles and practices will ensure that AI systems and tools used by the government, its contractors, vendors, and grantees are safe; do not perpetuate or exacerbate bias; would increase the responsible use of these systems in workplaces, and that AI systems be accountable to the American people. Agencies should pursue the creation and implementation of regulations needed to realize the goals of the AI Bill of Rights. New regulations would be in addition to enforcing existing civil rights protections.
For example, implementing the Algorithmic Discrimination Protections principle would mean federal agencies and contractors would need to ensure their automated systems provide protection from algorithmic discrimination, bias, or civil rights violations before those systems are funded, acquired, or put into use. As outlined in the AI Bill of Rights, these algorithmic protections should include:
- Use of representative data by developers and designers
- Independent algorithmic impact assessment and evaluation with results reported in plain language
- Equity assessments as part of the system design, as well as pre-deployment and ongoing disparity testing and mitigation
- Accessibility for people with disabilities in design and development
- Agencies’ and contractors’ clear organizational oversight of the use of AI tools and systems
Similar requirements should be developed for the other four principles in the AI Bill of Rights.
The new AI EO should make the AI Bill of Rights binding U.S. government policy and should lay out an aggressive timeline calling for the submission of agency implementation plans to the Executive Office of the President within 90 days and for full implementation within a year of the AI EO being issued. The AI EO should also establish a coordination mechanism, such as a new White House Council on AI, to coordinate executive and independent agencies for the implementation of the EO and on enforcement measures and resources where appropriate. Public accountability for these implementation plans is also necessary. Agencies already have a starting place for this effort with the public list of AI use case inventories required from each federal agency enacted by the previous administration’s EO 13960 and their OMB M-21-06 guidance,[7] but these AI use case inventories do not go far enough. The AI EO should require that White House coordination mechanism, OMB, and the National AI Initiative Office (NAIIO) to (1) expand the required questions for these AI use case inventories to include information that would allow the public to assess adherence with the AI Bill of Rights; (2) publish the inventories in a format that is more understandable and usable by the American people; and (3) require an annual report assessing each agency’s progress in implementing the AI Bill of Rights for their AI use cases.
The forthcoming OMB AI guidance alone is not sufficient to make the AI Bill of Rights binding U.S. government policy. Instead, the OMB AI guidance should provide detailed instructions to agencies on how to implement each of the five principles and practices of the AI Bill of Rights under a new AI EO.[8] Similarly, federal government and contractor compliance with the National Institute of Standards and Technology AI Risk Management Framework (NIST AI RMF) is not a substitute for making the AI Bill of Rights binding U.S. government policy via EO. While an AI EO should require all AI tools deployed by federal agencies or contractors to be assessed under the NIST AI RMF and summaries to be publicly released, such compliance is not a substitute for a people-first, rights-respecting approach; to accomplish this, the AI Bill of Rights must be made binding policy. Each of these actions is needed to help ensure effective accountability mechanisms are in place.
The AI Bill of Rights must be binding AI policy for the entire federal government.
The AI EO making the AI Bill of Rights binding must not exempt any part of the federal government, including law enforcement or the national security community. Law enforcement and judicial use of AI systems can have severe harmful consequences, especially for immigrants and communities of color, where those AI systems are biased and discriminatory.[9] The course the federal government sets on AI for national security and public safety will be closely watched by state and local governments as well as governments around the world. This requirement should also include federal contractors. Additionally, any federal grants used by state or local governments, including law enforcement agencies, to purchase automated systems must also ensure those systems comply with the AI Bill of Rights.
The federal government should lead on data privacy by adopting standards for data minimization for federal agencies and supporting federal privacy legislation.
Fully addressing the Data Privacy principle of the AI Bill of Rights will require the kind of federal privacy legislation many of our organizations have previously supported[10] and that the President has already called for in multiple State of the Union addresses.[11] However, the federal government can meet the Data Privacy principles of the AI Bill of Rights by having federal agencies and contractors adopt and meet new data minimization standards, such as those laid out in the bipartisan American Data Privacy and Protection Act (ADPPA) introduced in the 117th Congress.[12] That means prohibitions and limits placed on the collection of sensitive data, as well as limiting data collection generally to what is needed to provide the product or service.
Conclusion
We understand that details will need to be finalized on how to best implement the AI Bill of Rights for the federal government’s use of AI as well as to execute the forthcoming National AI Strategy and National AI Priorities. In the meantime, we applaud the recent actions from executive and independent agencies, including the recent joint statement from the Department of Justice, Federal Trade Commission, Consumer Financial Protection Bureau, and Equal Employment Opportunity Commission,[13] and know the administration will continue to ensure agencies mitigate AI harms in the workplace, to consumers, and beyond. We are also heartened to see a serious bipartisan effort emerge in the Senate to explore potential AI legislation and know that the administration will continue to work closely with Congress on these efforts.[14]
The forthcoming AI EO presents a clear opportunity to implement the White House’s own AI Bill of Rights. We urge you not to miss this critical chance to operationalize the values your administration has uplifted.
Thank you for your attention to these issues. For any questions or for further discussion, please contact Adam Conner, Vice President for Technology Policy, at [email protected], Anita Banerji, Senior Program Director, Media & Technology, at [email protected], or Samir Jain, Vice President of Policy, at [email protected].
Sincerely,
Center for American Progress
The Leadership Conference on Civil and Human Rights
Center for Democracy & Technology
Data & Society Research Institute
Lawyers’ Committee for Civil Rights Under Law
NAACP
Public Citizen
Public Knowledge
UnidosUS
cc:
Jeff Zients, Assistant to the President and Chief of Staff
Bruce Reed, Assistant to the President and Deputy Chief of Staff
Jake Sullivan, Assistant to the President for National Security Affairs, National Security Council
Lael Brainard, Assistant to the President and Director, National Economic Council
Neera Tanden, Assistant to the President and Domestic Policy Advisor, Domestic Policy Council
Arati Prabhakar, Assistant to the President for Science and Technology, Director of the White House Office of Science and Technology Policy
Shalanda Young, Director, Office of Management and Budget
[1] White House, “Biden-Harris Administration Secures Voluntary Commitments from Leading Artificial Intelligence Companies to Manage the Risks Posed by AI,” Fact Sheet, July 21, 2023, available at https://www.whitehouse.gov/briefing-room/statements-releases/2023/07/21/fact-sheet-biden-harris-administration-secures-voluntary-commitments-from-leading-artificial-intelligence-companies-to-manage-the-risks-posed-by-ai/.
[2] The White House Office of Science and Technology Policy, “Blueprint for an AI Bill of Rights,” (Washington: The White House, 2022), available at https://www.whitehouse.gov/ostp/ai-bill-of-rights/.
[3] White House, “Biden-Harris Administration Announces New Actions to Promote Responsible AI Innovation that Protects Americans’ Rights and Safety,” Fact Sheet, May 4, 2023, available at https://www.whitehouse.gov/briefing-room/statements-releases/2023/05/04/fact-sheet-biden-harris-administration-announces-new-actions-to-promote-responsible-ai-innovation-that-protects-americans-rights-and-safety/.
[4] The White House Office of Science and Technology Policy, “Blueprint for an AI Bill of Rights.”
[5] Suresh Venkatasubramanian, “The White House Already Knows How to Make AI Safer,” Wired, July 25, 2023, available at https://www.wired.com/story/the-white-house-already-knows-how-to-make-ai-safer/.
[6] Adam Conner, “The Needed Executive Actions to Address the Challenges of Artificial Intelligence,” Center for American Progress, April 25, 2023, available at
[7] Executive Office of the President, “Promoting the Use of Trustworthy Artificial Intelligence in the Federal Government,” Executive Order 13960, December 3, 2020, available at https://www.federalregister.gov/documents/2020/12/08/2020-27065/promoting-the-use-of-trustworthy-artificial-intelligence-in-the-federal-government; Russel T. Vought, “Memorandum for the Heads of Executive Departments and Agencies: Guidance for Regulation of Artificial Intelligence Applications,” Memorandum, Office of Management and Budget, November 17, 2020, available at https://www.whitehouse.gov/wp-content/uploads/2020/11/M-21-06.pdf.
[8] The Leadership Conference on Civil and Human Rights, “Next Steps to Advance Equity and Civil Rights in Artificial Intelligence and Technology Policy,” Letter, June 13, 2023, available at https://civilrights.org/resource/next-steps-to-advance-equity-and-civil-rights-in-artificial-intelligence-and-technology-policy/.
[9] Julia Angwin and others, “Machine Bias,” ProPublica, May 23, 2016, available at https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing; Cade Metz and Adam Satariano, “An Algorithm That Grants Freedom, or Takes It Away,” The New York Times, February 6, 2020, available at https://www.nytimes.com/2020/02/06/technology/predictive-algorithms-crime.html.
[10] The Leadership Conference on Civil and Human Rights, “Letter to House Energy and Commerce Committee on the American Data Privacy and Protection Act,” Letter, July 20, 2022, available at https://civilrights.org/resource/letter-to-house-energy-and-commerce-committee-on-the-american-data-privacy-and-protection-act/.
[11] The White House, “State of the Union Address,” March 1, 2022, available at https://www.whitehouse.gov/state-of-the-union-2022; The White House, “State of the Union Address,” February 7, 2023, available at https://www.whitehouse.gov/state-of-the-union-2023.
[12] American Data Privacy and Protection Act, H.R. 8152, 117th Congress, 2 sess. (December 30, 2022), available at https://www.congress.gov/bill/117th-congress/house-bill/8152/text.
[13] Federal Trade Commission, “Joint Statement on Enforcement Efforts Against Discrimination and Bias in Automated Systems,” Press Release, April 25, 2023, available at https://www.ftc.gov/system/files/ftc_gov/pdf/EEOC-CRT-FTC-CFPB-AI-Joint-Statement%28final%29.pdf.
[14] Senate Democrats, “Majority Leader Schumer Delivers Remarks To Launch SAFE Innovation Framework For Artificial Intelligence At CSIS, Speech, June 21, 2023, available at https://www.democrats.senate.gov/news/press-releases/majority-leader-schumer-delivers-remarks-to-launch-safe-innovation-framework-for-artificial-intelligence-at-csis#.