Biden signs Executive Order 14110 on AI

Background

An executive order is a directive from the US president to federal agencies. On 30 October 2023, President Biden signed Executive Order 14110, titled “Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence” ^[1]. The order was published in the Federal Register on 1 November 2023 at 88 FR 75191 ^[1].

What the order did

The order set out eight guiding principles and priorities for federal AI policy ^[1]. It did not assign a single task to a single agency; it directed work across many federal bodies. The Congressional Research Service (CRS) counts that the order directs over 50 federal entities to engage in more than 100 specific actions ^[2]. The CRS groups those actions into eight policy areas: safety and security, innovation and competition, worker support, civil rights, consumer protection, privacy, federal use of AI and international leadership ^[2].

Key definition

The order defines a “dual-use foundation model” as an AI model trained on broad data that generally uses self-supervision, contains at least tens of billions of parameters and is applicable across a wide range of contexts ^[1]. The definition identifies a class of large, general-purpose models rather than a single product or system. It also gives federal policy a shared term for models that can be used in many settings.

Labels and provenance

The order directed work on labelling and content provenance mechanisms so that Americans can determine when content is generated using AI ^[1]. Content provenance means information about how and where a piece of content was produced. Such mechanisms are intended to help people tell whether content came from an AI system. The order treats this as a labelling and provenance task, not only a technical research question ^[1].

Privacy

The order called for evaluation and mitigation of privacy risks associated with the collection, use and retention of user data ^[2]. Privacy is also one of the eight policy areas in the CRS grouping ^[2]. This places privacy alongside safety, civil rights and consumer protection in the order’s list of concerns.

Scope

The order combines principles, definitions and assigned actions. It sets eight guiding principles and priorities ^[1], defines a dual-use foundation model ^[1], and distributes work to over 50 federal entities through more than 100 specific actions ^[2]. Those actions fall into eight policy areas, from safety and security to international leadership ^[2]. The Federal Register text is the published version of the order, running at 88 FR 75191 ^[1]. The CRS count shows that the order relies on many agencies rather than a single office, and that it links AI policy to innovation, workers, civil rights, consumers, privacy and international leadership ^[2].