In a landmark move that signifies the growing importance of regulating artificial intelligence in the workplace, California has introduced new rules for Automated-Decision Systems (ADS) in employment. These regulations reshape the landscape for businesses using AI tools to make hiring and employment decisions, marking California’s commitment to leading in this domain of innovation and ethical governance. The decision by the California Civil Rights Council to approve the Employment Regulations Regarding Automated-Decision Systems sets a new precedent for the use of technology in employment practices. The focus is on ensuring these tools do not discriminate against individuals and that businesses maintain transparent evidence of compliance. As these regulations are now implemented, employers across California need to closely examine their existing AI systems and prepare for a future where compliance and fairness go hand in hand.
Understanding the Scope of Automated-Decision Systems
The new regulation defines an Automated-Decision System as any computational process that makes or aids an employment-related decision. This scope is extensive, encompassing systems utilizing AI, machine learning, algorithms, statistics, or any similar data-processing techniques. Examples include, but are not limited to, tools for resume screening, the ranking of job applicants, tailored job advertising, and analysis of interviewees’ facial expressions or tone. Moreover, systems that evaluate personality, aptitude, reaction time, or cultural compatibility also fall under this regulation. It is notable that general-purpose technologies like spreadsheets or cybersecurity software are excluded, except when they support employment decisions directly. A significant aspect of the regulation is the lack of a strict definition of what constitutes “facilitation,” which provides room for interpretation. For instance, a tool that highlights discrepancies in an applicant’s educational background could influence hiring outcomes without directly making a rejection. Employers are now tasked with assessing whether their systems facilitate decision-making in areas that necessitate compliance. In essence, businesses must conduct a detailed analysis of how integral such technologies are to their processes and whether these uses trigger obligations under the new law.
Implications for Employers and Agents
The regulation extends its reach to include agents acting on behalf of employers under the Fair Employment and Housing Act (FEHA). An agent, in this context, refers to any individual or organization performing duties traditionally done by employers, potentially using Automated-Decision Systems. This broad definition indicates that all third-party vendors involved in activities like recruitment, promotions, or compensation decisions fall within this purview. Although most employers might not conduct in-house background checks, even external vendors can be considered agents if their services meaningfully influence hiring decisions. While background screening vendors might not fit the definition of an “employment agency” under this law, businesses should explore whether these services act as facilitators in their hiring processes. Importantly, an employment agency is defined as any entity that, for compensation, sources job applicants, employees, or work opportunities, which now includes the use of Automated-Decision Systems. Employers must be keenly aware of the responsibilities these definitions impose and be prepared to hold third-party service providers accountable.
Criminal History and Recordkeeping Considerations
These rules reinforce California’s Fair Chance Act by prohibiting the use of ADS to evaluate an applicant’s criminal history before a conditional job offer. Should an offer be retracted, the decision must be based on an individualized judgment, detailing why the criminal record was the reason for this retraction. This process must hold regardless of whether the decision is automated or involves human intervention. While an earlier draft of the regulation required businesses to provide applicants with reports generated by ADS, this clause was omitted in the final version. Nevertheless, businesses are encouraged to explain how ADS contributed to their decision-making process and to ensure adherence to required procedural protections.
Moreover, the final regulation requires employers to maintain detailed personnel records and ADS data for a minimum of four years. This directive extends to all data that contributes to developing or customizing the system, including inputs, outputs, and data about individual applicants or employees. Though the regulation does not make anti-bias testing mandatory, evidence of such testing—or lack thereof—plays a critical role in determining liability. Consequently, organizations that regularly conduct and document such testing could better defend against discrimination claims.
Vendor Relationships and Compliance Strategies
The regulations necessitate a thorough examination of employer and vendor relationships to ensure compliance. While third-party vendors or developers of ADS tools are not directly held accountable under these rules, employers using such technology assume responsibility for any discrimination caused by these tools. Hence, businesses must take proactive steps to understand how ADS tools are constructed, trained, and applied. This involves acquiring comprehensive documentation on the tool’s purpose, design, and testing outcomes, alongside establishing clear roles and obligations through contractual agreements. Fulfilling these responsibilities helps ensure that the ADS employed in decision-making align with California law and broader ethical standards.
For businesses employing technologies that make or facilitate employment decisions, demonstrating compliance necessitates transparency. Employers must meticulously document their processes and decision-making criteria to show that their use of Automated-Decision Systems meets regulatory requirements. Especially critical is a focus on tools used in sensitive decisions—those that might be based on personality assessments, facial recognition, or criminal history evaluations. This detailed comprehension and transparent documentation are essential to addressing potential challenges of discrimination or bias.
Key Actions for Employers
To effectively navigate the new regulations, employers are encouraged to undertake several key actions. First, conducting an inventory of all technological tools used in employment activities is crucial. This inventory should assess whether the technology fits within the ADS definition. Additionally, businesses must meticulously review vendor relationships to determine if service providers classify as agents and whether their tools introduce legal risks. Developing an internal governance program stands out as another critical strategy. This program should incorporate comprehensive documentation, potential anti-bias testing protocols, and a recordkeeping process that aligns with the four-year requirement. Evaluating high-risk use cases is also imperative, focusing on technologies built upon personality assessments, facial recognition, or criminal history analysis. Such in-depth assessments will prepare organizations to handle potential challenges effectively while fostering a culture of compliance and ethical use of technology.
Path Forward for Organizations
The new regulation defines an Automated-Decision System as any computational tool making or assisting employment decisions, including those utilizing AI, machine learning, statistics, or related techniques. This broad definition covers tools for resume screening, ranking job candidates, targeted job ads, and assessing interviewees’ facial expressions or tone. It also includes systems evaluating personality, aptitude, reaction time, or cultural fit. However, general-purpose technologies like spreadsheets or cybersecurity software are excluded unless they explicitly support employment decisions.
A notable aspect of this regulation is its vague definition of “facilitation,” leaving room for interpretation. For example, a tool that flags discrepancies in an applicant’s education may influence hiring decisions even without directly rejecting a candidate. Employers must now evaluate whether their systems aid decision-making in areas requiring compliance. Essentially, businesses are tasked to carefully analyze how essential these technologies are to their operations and if their use prompts obligations under the new law.