Summer of AI Action
Between May and August 2023, the White House has announced multiple actions and executive orders related to protecting Americans from potential risks of AI, while working to maximize the U.S. leadership in AI innovation. Key announcements and actions, include:
- 4 May 2023: (Announcement) White House Announces New Actions to Promote Responsible AI Innovation that Protects Americans’ Rights and Safety
- Vice President Harris meets with CEOs from Alphabet, Anthropic, Microsoft, and OpenAI and announces:
- New investments in AI research and development (R&D) via the National Science Foundation for seven new AI Research Institutes (18 → 25 such Institutes);
- Public red-teaming of existing genAI systems at DEFCON in August ; and
- Office of Management and Budget Draft Policy Guidance on use of AI systems by U.S. federal agencies due later in the summer.
- Vice President Harris meets with CEOs from Alphabet, Anthropic, Microsoft, and OpenAI and announces:
It was exciting to see so much interest in AI and in the red-teaming exercise at DEFCON. Many people waited in line for over an hour just to get in the room. There was so much interest that the systems running the competition could not keep up with demand and crashed periodically. In the future, similar exercises can be a great way, not only to red-team some of these models, but also to educate the public on some of the risks and harms associated with large language models. Since users can interact with them entirely in natural language, the barrier to entry is relatively low, and users do not need to be a professional hacker or an AI expert to participate.Jessica Ji, CSET Research Analyst
- 23 May 2023: (Announcement) White House Takes New Steps to Advance Responsible AI Research, Development, and Deployment
- White House announces updated National AI R&D Strategic Plan, released by the Office of Science and Technology Policy (OSTP), which outlines priorities and goals for AI R&D investments by the federal government. The announcement also includes an OSTP Request for Information (RFI) on national priorities regarding AI Risks.
- Related CSET Work:
- RFI Response on National Artificial Intelligence Research and Development Strategic Plan (March 2022)
- 21 July 2023: (Announcement) White House Announces Voluntary Commitments from Leading AI Companies to Manage Risks Posed by AI
- Seven leading companies including Anthropic and OpenAI commit to ensure safe products before public release, building AI systems that put (cyber)security first, and earning the public’s trust to enable the public to know when content is AI generated.
- Related CSET Work:
- Blog Post: Making AI (more) Safe, Secure, and Transparent: Context and Research from CSET (21 July 2023)
- 9 August 2023: (Executive Order) Executive Order on Addressing United States Investments in Certain National Security Technologies and Products in Countries of Concern
- The Executive Order (EO) and associated Treasury rule-making (ANPRM) seeks to curb U.S. investments in key technology sectors in China–namely semiconductors and microelectronics.
- Related CSET Work:
- Blog Post: Scoping AI For National Security: An Impossible Task? (28 August 2023)
- Report: U.S. Outbound Investment into Chinese AI Companies (February 2023)
- Commentary in Barrons: The U.S. Wants to Target China’s Dangerous AI Systems. It’s Easier Said Than Done (18 August 2023)
- 9 August 2023: (Announcement) White House Launches AI Cyber Challenge to Protect America’s Critical Software
- DARPA competition partnering with Anthropic, Google, Microsoft, and OpenAI on AI Cyber Challenge to “challenge competitors across the United States, to identify and fix software vulnerabilities using AI”
- Some additional information shared at DEFCon/BlackHat Conference on 9 August 2023
- Two tracks: Open and Funded. Competition qualifying events will occur in Spring 2024 (April) with semifinals occurring at DEFCON 2024 (August). Finals at DEFCON 2025 (August 2025).
- Related CSET Work:
- Report: Federal Prize Competitions (November 2021)
- DARPA competition partnering with Anthropic, Google, Microsoft, and OpenAI on AI Cyber Challenge to “challenge competitors across the United States, to identify and fix software vulnerabilities using AI”
Earlier AI Actions
These recent actions and announcements add to work undertaken earlier in the Biden administration and in the Trump administration to capitalize on AI development and protect the American people from potential harms. Prior actions, in reverse chronological order, include:
- 16 February 2023: (Executive Order, 14091) Executive Order on Further Advancing Racial Equity and Support for Underserved Communities Through The Federal Government
- Provisions here build upon prior EOs, including EO13895, to extend and strengthen equity-advancing requirements for agendas. For example, this EO supports “promot(ing) equity in science and root(ing) out bias in the design and use of new technologies, such as artificial intelligence.”
- 26 January 2023: (Other Agency, Announcement) National Institute of Standards and Technology (NIST) Releases AI Risk Management Framework v.1.0
- Framework developed in collaboration with public and private sectors to “better manage risks to individuals, organizations, and society associated with AI.”
- Related CSET Work:
- RFI Response: Comment to NIST on the AI Risk Management Framework (29 September 2022)
- Blog Post: Translating AI Risk Management Into Practice (26 June 2023)
- 4 October 2022: (Announcement) White House Announces Key Actions to Advance Tech Accountability and Protect the Rights of the American Public (AI Bill of Rights Blueprint)
- OSTP releases Blueprint for an AI “Bill of Rights” to help “guide the design, development, and deployment of AI and other automated systems.” The Bill of Rights lays out five core protections:
- Safe and Effective Systems;
- Algorithmic Discrimination Protections;
- Data Privacy;
- Notice and Explanation; and
- Alternative Options.
- The Blueprint follows extensive public comment consultations that began in November 2021.
- OSTP releases Blueprint for an AI “Bill of Rights” to help “guide the design, development, and deployment of AI and other automated systems.” The Bill of Rights lays out five core protections:
- 8 December 2020: (Executive Order, 13960) Executive Order on Promoting the Use of Trustworthy Artificial Intelligence in the Federal Government
- EO signed by President Trump stating that agencies should be transparent in disclosing information regarding their use of AI to appropriate stakeholders.
- The EO directs agencies to conduct annual inventories of their AI use cases, and publish them whenever possible starting in June 2022.
- 11 February 2019: (Executive Order, 13859) Executive Order on Maintaining American Leadership in Artificial Intelligence
- EO creates the “American AI Initiative” to spur development and regulation of AI, machine learning, and deep learning and calls on OMB to provide guidance to federal agencies on the regulation of AI applications.