Artificial intelligence (“AI”) software programs are generating enormous legal concerns around the world. The European Union has recently promulgated significant regulations intended to reduce various risks related to AI programs. Small-to-large AI programs are used for a large number of services and are embedded in devices that might be surprising to consumers. Examples include:

• Energy production, generation, and delivery
• Public and private transportation systems like trains, subways, airplanes, and more
• Law enforcement services at every level, including monitoring and preventing criminal behavior
• Banking
• Government provision of services
• Military applications at “high levels” and embedded within weapons systems
• Corporate use in many aspects including employment hiring and retention
• Artistic endeavors like films, television, books, music and other media
• Medical devices and provision of health services

The various risks relate to whether the AI program might “go rogue,” might be subject to misuse through accident, mistake, or hacking, might create societal or individual harm through malfunction, defect, or cessation of functioning, might create harm because inputs might be false or biased, etc., or might be harmful though lack of human input and supervision.

In the United States, there have been some sporadic efforts to address some of these potential harms and risks. For example, New York — and other States — have passed laws attempting to regulate how AI programs are used by businesses in making hiring decisions.

California has added a new element to the ongoing discussion of the risks that AI programs might pose to society and individuals. California recently passed Senate Bill 1047, which is commonly called the California Killswitch law but is officially called Safe and the “Secure Innovation for Frontier Artificial Intelligence Models Act.” See the text of the bill here. The general idea is that large AI programs must add a “killswitch.”

If signed by Governor Newsom, the legislation will apply to any AI model with the following characteristics:

(i) An artificial intelligence model trained using a quantity of computing power greater than 10^26 integer or floating-point operations, the cost of which exceeds one hundred million dollars ($100,000,000) when calculated using the average market prices of cloud compute at the start of training as reasonably assessed by the developer.

(ii) An artificial intelligence model created by fine-tuning a covered model using a quantity of computing power equal to or greater than three times 10^25 integer or floating-point operations.

The various dollar amounts and computing power are to be revised upward based on inflation and other technological developments.

The harms and risks that are being addressed by the legislation are these:

(A) The creation or use of a chemical, biological, radiological, or nuclear weapon in a manner that results in mass casualties.

(B) Mass casualties or at least five hundred million dollars ($500,000,000) of damage resulting from cyberattacks on critical infrastructure by a model providing precise instructions for conducting a cyberattack or series of cyberattacks on critical infrastructure.

(C) Mass casualties or at least five hundred million dollars ($500,000,000) of damage resulting from an artificial intelligence model engaging in conduct that does both of the following: (i) Acts with limited human oversight, intervention, or supervision and (ii) Results in death, great bodily injury, property damage, or property loss, and would, if committed by a human, constitute a crime specified in the Penal Code that requires intent, recklessness, or gross negligence, or the solicitation or aiding and abetting of such a crime.

(D) Other grave harms to public safety and security that are of comparable severity to the harms described in subparagraphs (A) to (C), inclusive.

To lessen the risks of these potential harms, SB 1047 requires that developers of covered AI programs implement various policies and procedures (like cybersecurity) and also provide a “killswitch” for the AI program. Specifically, the legislation states in Section 22603: “Before a developer initially trains a covered model, the developer shall do all of the following … (2) Implement the capability to promptly enact a full shutdown.”

The legislation is awaiting the Governor’s signature. If signed, a new era of AI regulation will begin. Even if vetoed, the concepts and fears underlying the legislation will continue to resonate with regulators. It is noted that the European Union’s new regulations do not contain any provisions for requiring a “killswitch.” Will such a requirement be added?

Contact the AI and Internet Attorneys at Revision Legal

For more information, contact the experienced AI and Internet Lawyers at Revision Legal. You can contact us through the form on this page or call (855) 473-8474.