OSTP RFI - National AI Strategy

On July 7, 2023, CAIDP sent detailed comments to the Office of Science and Technology Policy regarding the US National AI Strategy. A few of the key points:

• "Legal standards should be established to ensure that AI systems are designed, developed, and deployed in a manner that protects people’s rights and safety. Without legal standards, accountability mechanisms are lacking, and trust and safety practices cannot be enforced. Companies should not release AI products that are not safe."

• "We also recommend prohibitions on the deployment of certain AI systems. For example, pseudo-scientific AI systems such as mass facial surveillance, predictive policing, emotion recognition and biometric categorization should be prohibited."

• "Assigning legal liability to upstream actors is necessary to ensure the quality, safety, and robustness of foundation models and to monitor and mitigate possible biases and harms. A clear liability regime would also provide transparency and protection for the users and applications downstream. "

• "Lethal autonomous weapons pose the greatest risk to national security. The United States should prohibit such systems and champion moratoria on their development. "

• "The data used to train and test an AI model should be representative of the demographic groups that will interact with the model in its final form."

• "Promoting algorithmic transparency will also help mitigate algorithmic bias. Individuals must have the ability to meaningfully contest adverse decisions and to access the logic, data, and functions that contributed to the outcome. "

• "AI systems pose a set of unique challenges to democratic processes. Misinformation, disinformation, and deepfakes powered by AI can distort public discourse, influence public opinion, and potentially interfere with electoral processes. Microtargeting of political messages, powered by AI, can also lead to increased polarization and manipulation of voters."

CAIDPs Recommendations

1) Ensure the development of human-centered and trustworthy Artificial Intelligence based on fundamental rights, democratic values, and the rule of law 

2) Prioritize investment in AI systems that are innovative and ensure public safety 

3) Establish guardrails for AI based on transparency, contestability, traceability, robustness, safety, security and accountability. 

4) Implement the OSTP AI Bill of Rights, the OECD AI Principles, and the UNESCO Recommendations on AI Ethics

The Center for AI and Digital Policy supports the Request for Information concerning a National AI Strategy, announced by the Office of Science and Technology Policy on May 23, 2023. We intend to submit comments. In advance of the deadline, we offer several recommendations to commentators.

• Read carefully the Request for Information. Understand the context for the RFI and the type of information that the Office of Science and Technology Policy is seeking from the public
• Learn about OSTP's prior work on AI policy. The Office of Science and Technology Policy has already launched several important AI policy initiatives, including the Blueprint for an AI Bill of Rights. You may be able to reference this earlier work in support of your recommendations.
• Answer the questions that are most relevant to your work and expertise. It is not necessary to answer all of the questions
• Write clearly and directly. Short, simple statements are often the most effective.
• Provide evidence. If you make a factual claim, include a citation to a report, article, or study to support your point.
• Make concrete recommendations. If you have a specific recommendation for OSTP, state it.
• Keep a copy of your comments. Managing the federal portal for agency comments is not easy. The best strategy is often to prepare your comments in advance and then cut and past the sections into the form,
• Evaluate the outcome. You were asked by a public agency for your views. You took time to prepare a response. You should expect the agency to consider your comments. The agency may not agree with you, but you are entitled to a "reasoned response" to your recommendations.

Good luck!

CAIDP has made several recommendations for the US National AI Strategy:

• Enactment of federal legislation for the governance of AI
• Implementation of the OSTP AI Bill of Rights
• Implementation of the OECD AI Principles (which the United States has already endorsed)
• Support for the Council of Europe AI Treaty (a global binding treaty)

• CAIDP, OSTP AI Bill of Rights
• Merve Hickok,  Advances in AI: Are we ready for the tech revolution, Testimony Before the House Oversight Committee (March 8, 2023)
• Lorraine Kisselburgh and Marc Rotenberg, Next Steps on the AI Bill Of Rights, Washington Spectator (Nov. 2021)
• Marc Rotenberg and Merve Hickok, Regulating A.I.: The U.S. Needs to Act,The New York Times (March 6, 2023)
• Marc Rotenberg and Merve Hickok, Artificial Intelligence and Democratic Values: Next Steps for the United States,  Council on Foreign Relations (August 22, 2022)

[This is an excerpt from the US country report, prepared by CAIDP]

The U.S. lacks a unified national policy on AI but President Biden, and his top advisors, has expressed support for AI aligned with democratic values. The United States has endorsed the OECD/G20 AI Principles. The White House has issued two Executive Orders on AI that reflect democratic values, a federal directive encourages agencies to adopt safeguards for AI. The most recent Executive Order also establishes a process for public participation in the development of federal regulations on AI though the rulemaking has yet to occur. The overall U.S. policy-making process remains opaque and the Federal Trade Commission has failed to act on several pending complaints concerning the deployment of AI techniques in the commercial sector. But the administration has launched new initiatives and encouraged the OSTP, NIST, and other agencies to gather public input. The recent release of the Blueprint for an AI Bill of Rights by the OSTP represents a significant step forward in the adoption of a National AI Policy and in the U.S.’s commitment to implement the OECD AI Principles. There is growing opposition to the use of facial recognition, and both Facebook and the IRS have cancelled facial recognition systems, following widespread protests. But concerns remain about the use of facial surveillance technology across the federal agencies by such U.S. companies as Clearview AI. The absence of a legal framework to implement AI safeguards and a federal agency to safeguard privacy also raises concerns about the ability of the U.S. to monitor AI practices.

[More information about the AI and Democratic Values Index]