AIVI Act: Illinois Leading the Way on AI RegulationPosted on 9th September 2019
Illinois, known as a national test bed for regulation, has struck again in an unlikely area; the use of artificial intelligence (AI) in job interviews. On 11 August, Democratic Governor JB Pritzker signed the Illinois Artificial Intelligence Video Interview Act (AIVI Act) into law. Illinois is now the first state to regulate the use of algorithms (interview bots) and other forms of AI to analyse applicants during video interviews from the beginning of 2020.
AI in video interviews is a growing trend in hiring practices across the US. Companies rely on video interviews to reduce the stages of these processes to get a sense for candidates, who are asked to answer questions and present themselves just as they would in an in-person interview. Companies typically use the services of third-party platforms such as Hire Vue to conduct video interviews. Such platforms are able to provide an analysis of applicants’ facial expressions, body language, choice of words and vocal tones. With the passage of the AIVI Act, there are certain obligations, disclosures and uncertainties that companies, likely to utilise these types of services, will face.
The AIVI Act
In accordance with the AIVI Act, employers intending to conduct AI video interviews, when hiring for positions based in Illinois, must satisfy several conditions. Employers must provide advanced disclosures to applicants if they intend to utilise AI to analyse their video interviews. Second, an explanation must be given to applicants on how the technology works and how it will assess them. Applicants must consent to the use of AI before it is used to analyse video interviews and employers must not share the video with anyone other than those responsible for deciding whether a candidate is suitable for the job. Moreover, all original and back-up copies of the interview must be destroyed by employers within 30 days of receiving a request from the applicant.
AIVI Act Omissions
First, the Act does not define “artificial intelligence”. This has been problematic for companies globally, as AI has been defined differently across various technologies that involve learning, evolution and analysis. Given the lack of clarity as to the definition of AI under the legislation, companies may not determine that their video interview tools, and analysis constitute AI technology under the Act.
The Act also does not define the information that must be provided to candidates during required disclosures. This lack of scope may be intentional to drive companies to be as transparent as possible with applicants, however, it also creates a level of uncertainty for companies and raises questions as to the threshold of sufficient information.
Another broad and challenging provision, likely to cause uncertainty for companies, is the lack of determination on how consent from applicants should be given. Methods of establishing consent could range from written statements to a click wrap agreement, but the Act itself does not specify what constitutes consent.
Finally, the Act does not establish penalties for violations or breaches, nor does it describe the support and solutions that may be available, or quantify damages. Therefore, it leaves the question of how its provisions will be enforced open to interpretation. Initially, this is to be decided on a case-by-case basis until precedents have been established.
Concerns and Opportunities
The AIVI Act in Illinois will likely run parallel to the Illinois Biometric Information Privacy Act (BIPA), which requires disclosure and consent before businesses can collect and use biometric information. This includes voice and facial recognition software, as well as other biometric information for identification and security purposes. If Illinois businesses use AI in video interviews and the AI employs facial analysis technologies, vocal identifiers or other biometric information covered by BIPA, their violation of disclosure and consent requirements of the AIVI Act will also violate BIPA. The AIVI Act’s deletion of data requirements will also conflict with BIPA’s deletion provisions. BIPA requires biometric data to be deleted within three years of the last interaction with a person, while the AIVI Act only requires deletion if requested by an applicant. Thus, although the video data contains biometric identifiers, the AIVI Act does not require data to be deleted like BIPA would.
Other issues businesses will contend with include: How does a company explain what an algorithm does to applicants? What happens when a candidate does not consent to the use of AI to evaluate their video interview? When an applicant asks to delete their data, how does that conflict with federal laws that demand the preservation of evidence or other document retention requirements?
Companies that comply with Europe’s General Data Protection Regulation (GDPR) are ahead. At an Illinois Chamber of Commerce event, Director of Data Science at Hire Vue, Lindsey Zuloaga, stated Hire Vue is set to comply with the Illinois law due to their ability to already meet compliance requirements under GDPR.
Although the AIVI Act raises numerous concerns, it also encourages the use of emerging technologies like AI. By setting parameters and promoting information sharing, the Act increases the acceptability of AI by providing some protection for candidates and promoting the adoption of advanced technologies in the hiring process. It also plays a part in familiarising and normalising the use of emerging and advanced technologies in all aspects of life.
Could the AIVI Act Influence Legislation Federally?
Other US states will look to the Illinois AIVI Act as a model to regulate the use of AI during video interviews – but will it influence federal legislation? We do not believe it will in this Congress. While there is proposed legislation that requires assessment algorithms to evaluate their impacts on bias and other factors (Senator Ron Wyden’s D-OR Algorithmic Accountability Act), as well as a draft bill that proposes the creation of an AI regulatory agency that has the authority to block “misuses” of technology (Artificial Intelligence Data Protection Act), it is unlikely these bills will pass into law under the current Congress. Before the Federal Government addresses the AI question, it will likely focus on developing a federal comprehensive privacy law, however AI regulation is clearly on the horizon.
Author: Filip Pacyna, Policy Analyst, Access PartnershipBack to document archive