Gastbeitrag: Artificial Intelligence Legislation - An American Approach

In March, the European Union enacted the AI Act – a major milestone in artificial intelligence (“AI”) regulation.¹ Currently, there is no similarly comprehensive federal legislation or regulation on artificial intelligence in the United States. However, AI legislation – particularly in regard to the increasingly prevalent generative artificial intelligence – remains top of mind, frequently appearing on state and federal dockets and as a fixture within the Biden administration.

Outside of the federal system, AI legislation has been proposed, debated, and – in some cases – enacted across the United States. Most of this activity has occurred at the state level, though in certain instances there has been legislation at the city level, such as in New York.² This legislation has been met with mixed success. At the end of last year, seventeen states had enacted twenty-nine bills focused on AI regulation; most of these were focused on consumer protection and data privacy.³ While many of these regulations are in blue states – such as California, Illinois, New York, and Vermont – red states are also passing legislation, including Iowa, Louisiana, Tennessee, and Virginia.⁴ Many more states have proposed legislation that is yet to be enacted.⁵ With that said, an accurate catalogue of state and city AI ecosystem would be challenging and lengthy. As such, this report will primarily focus on activity at the federal level. Additionally, fifteen companies – including Amazon, Google, Meta, and Microsoft – have given commitments to the federal government to engage in safe AI practices.⁶ Similarly, twenty-eight healthcare providers and payers have made similar commitments in the healthcare space. Other companies and organizations have made similar public statements.⁷ While these non-governmental commitments are important, they are voluntary and non-binding.

Considering the dearth of explicit legislation regulating artificial intelligence, the federal government has found creative ways to regulate artificial intelligence, using existing legislation to fill in the gaps. As an example, the Telephone Consumer Protection Act – passed in 1991 – includes restrictions on the use of “artificial or pre-recorded voice messages,” which the Federal Communications Commission has ruled includes AI-generated human voices.⁸ Existing regulation can also be used for the opposite effect. For example, AI often falls into the category of “Export Administrative Regulations,” which regulates technology such as circuits and semiconductors, and at present is very lightly controlled.⁹ Others argue that AI would be more properly categorized as “Traffic in Arms” or “Munitions.”¹⁰ While change has been slow, in the interim there has been legislation which mentions artificial intelligence in part, even if not the focus of the legislation. One such example is the Federal Aviation Administration Reauthorization Act, which includes language requiring review of AI in aviation; another is the National Defense Authorization Act for Fiscal Year 2019, which provided for various AI-undertakings within the Department of Defense, including the appointment of an AI activities coordinator.¹¹ Additionally, regulatory bodies continue to hold that they control the space – for example, in April 2023, the Federal Trade Commission, Equal Employment Opportunity Commission, Consumer Financial Protection Bureau and Department of Justice issued a joint statement holding their authority applies to "software and algorithmic processes, including AI.”¹² These intermediary steps have been crucial to providing a workable framework for legislating AI.

In Congress, artificial intelligence legislation looms large; both the house and senate have passed bills attempting to regulate AI. Notable examples include the National Artificial Intelligence Initiative Act of 2020 (H.R.6216), establishing an American AI Initiative and providing guidance on AI research and development at federal science agencies¹³, and the AI in Government Act (H.R.2575), which created an “AI Center of Excellence.”¹⁴ The 117th Congress introduced at least seventy-five bills focused on AI, machine learning, or related issues, six of which were passed.¹⁵ The following congress introduced forty AI bills, none of which had passed as of last June.¹⁶ While these bills are not binding, they do point to the emphasis elected officials are placing on artificial intelligence regulation, and glimpse into the general temperature of congress regarding AI initiatives.

While not law, the closest document the United States has to the European Union’s AI Act is President Biden’s proposed AI Bill of Rights. The AI Bill of Rights is a set of five principles and related best practices with the goal of “[guiding] the design, use, and deployment of automated systems to protect the rights of the American public in the age of artificial intelligence.”¹⁷ The framework emphasizes alignment with “democratic values” and the protection of “civil rights, civil liberties, and privacy.”¹⁸ Specifically, it highlights five principles: Safe and Effective Systems, Algorithmic Discrimination Protections, Data Privacy, Notice and Explanation, and Human Alternatives, Consideration, and Fallback.¹⁹ The AI Bill of Rights serves as a cross-sector “national values statement and toolkit” meant to inform AI policy, practice, and protections.²⁰ Though not binding, it gives insight into President Biden and his administration’s approach to artificial intelligence, and perhaps even foreshadowing future AI legislation.

Currently, the most comprehensive piece of governance²¹ regarding artificial intelligence in the United States is President Biden’s recent executive order, Executive Order 14110²². Signed on October 30th, 2023, the executive order – also titled Executive Order on Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence – outlines Biden’s AI policy goals and orders executive agencies to act pursuantly. The goals fall into a select few buckets – safety, security, and other protections (“Ensuring the Safety and Security of AI Technology,” “Protecting Consumers, Patients, Passengers, and Students,” “Protecting Privacy”), human rights (“Supporting Workers,” “Advancing Equity and Civil Rights”), and progress (“Promoting Innovation and Competition,” “Advancing Federal Government Use of AI,” “Strengthening American Leadership Abroad”). To implement these goals, the order establishes the White House Artificial Intelligence Council within the Executive Office of the President and requires a number of federal agencies to appoint a chief artificial intelligence officer. In many cases, the agencies are also required to create additional guidelines for their sector or the industries they oversee. Most of the requirements must be completed by October 29th, 2024.²³ The reception was generally positive, with the mandates being referred to as “ambitious” and “comprehensive.”²⁴

It is unsettled if the United States’s more permissive approach helps or cripples the nation when compared to its European Union counterparts. While the EU has taken a precautionary approach to artificial intelligence, champions in the United States laud the country’s decentralized system.²⁵ While the US’s system provides fewer federal protections, it also allows for more state-specific flexibility, allowing states to react to their particular industries and sectors. However, as artificial intelligence develops it remains to be seen if the United States will continue with its current legislative model or if it will eventually develop a more comprehensive schema more analogous to the EU’s AI Act.

Surveying the legal landscape of artificial intelligence, it is undeniable there is an appetite for protection and regulation. Much of the current legislation has been steeped in cross-aisle collaboration, with artificial intelligence serving as a rallying point for Democrats and Republicans alike.²⁶ However, it is unclear if this trend will continue in the wake of impending election, which is one of the most contentious in the United States’s history. The results of November’s election will surely have ramifications, though the extent to which artificial intelligence will be featured at the ballot box is uncertain. Experts are anticipating major upheavals with the current artificial intelligence schema, regardless of which party wins the presidency.²⁷ In this tenuous political climate, one thing is evident – artificial intelligence will remain a priority for the foreseeable future.