Plain English Breakdown
The official status indicates the bill was ordered to an inactive file at the request of Senator Wiener, meaning it may not become law despite passing both chambers. The effective date is listed as blank in the metadata.
Rules for Automated Decision Systems
This law regulates computer systems using AI or data analytics that help replace human judgment in decisions affecting people, requiring testing and giving individuals a way to appeal those decisions.
What This Bill Does
- Defines an automated decision system as a computational process using machine learning, statistical modeling, data analytics, or artificial intelligence that issues simplified output like scores or recommendations to assist or replace human decision-making in ways that materially impact people.
- Requires developers of these systems to conduct performance evaluations and impact assessments before transferring them to users.
- Mandates that starting January 1, 2027, organizations must tell individuals when a system made a consequential decision about them and allow those individuals an opportunity to appeal the result as prescribed by future rules.
- Orders developers, deployers, or auditors to give unedited copies of their test results and impact assessments to the Attorney General within 30 days if asked.
- Keeps these specific reports private so they cannot be requested under public records laws.
Who It Names or Affects
- Developers who build automated decision systems used for consequential decisions.
- Organizations or agencies that use these systems to make choices about people.
- Individuals who are the subject of a consequential decision made by an automated system.
- The California Attorney General, who can request reports and file civil lawsuits against those who break the rules.
Terms To Know
- Automated Decision System (ADS)
- A computational process using machine learning, statistical modeling, data analytics, or artificial intelligence that issues simplified output like a score, classification, or recommendation to assist or replace human decision-making.
- Consequential decision
- A decision made by the system that materially impacts natural persons.
- Deployer
- The organization or agency that uses an automated decision system to make decisions about people.
Limits and Unknowns
- The law does not specify exactly which types of decisions count as 'consequential' beyond stating they must materially impact natural persons.
- Details on how appeals and disclosures must be handled are described only as being done in a way that is prescribed by future rules or regulations.