White House Issues Executive Order on Safe, Secure, Trustworthy AI

Person with glasses looks at digital screen

November 27, 2023

President Biden's executive order addresses safety, security, transparency, and other concerns surrounding the rapid growth of artificial intelligence 

On Oct. 30, the Biden administration released a wide-ranging executive order on Safe, Secure and Trustworthy AI. The executive order follows on policy directives and guidelines emerging across the globe aiming to harness the potentially transformative power of AI while mitigating its potentially dangerous risks related to user privacy, safety, and trust.

U.S. joins a global trend towards regulation of artificial intelligence

Since 2018, the European Union has published multiple communications and white papers on their approach to governing artificial intelligence and introduced their Artificial Intelligence Act in 2021. Canada introduced their own draft regulatory framework in 2022 in their Artificial Intelligence and Data Act (AIDA). Both frameworks apply risk-based approaches to assessing AI usage and safety risks, with the EU's AI Act going so far as to ban AI systems that are considered a threat carrying unacceptable risk.

The Biden administration has previously worked with the National Institute of Standards and Technology (NIST) on the safety of AI systems, and in 2022, the White House Office of Science Technology and Policy released a "Blueprint for an AI Bill of Rights." Although guidelines and standards have been slowly emerging in specific industries, the new executive order is set to impact a much wider range of domains, spanning from consumer products, housing and basic needs, transportation (including autonomous vehicles), through healthcare. 

A broad definition of artificial intelligence that cuts across applications

The Biden administration's executive order is very broad, based in part on the definition of artificial intelligence as "a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments." This definition signals impact for algorithms of varying levels of sophistication and architecture types, from supervised machine-learning predictive AI through generative AI. 

The president charges the secretary of defense, the secretary of commerce, NIST, and others to establish norms, policies, and initiatives around AI, in some cases tasking specific deadlines from 90 to 365 days from the date of the order. This sets the stage for evolving changes to the regulatory landscape in the near and long term.

While the initiatives set forth within the executive order itself will primarily determine the implementation details, the order provides eight guiding principles, such as planning for safety and security, equity and civil rights, privacy, and protecting end users of AI and AI-enabled products across industries.

What Can We Help You Solve?

Exponent can conduct a holistic assessment of your AI pipeline, including data collection, model training, validation protocols, deployment, post-deployment behavior, and user engagement, as well as intended and unintended use cases to improve performance and mitigate safety risks. We combine these efforts with user research and statistical modeling to identify vulnerable demographics, user subgroups, and safety-critical scenarios.