TL;DR: As of January 1, 2026, California companies must begin "risk assessments" for automated decision-making technology (ADMT). This is the first step in a major regulatory rollout. By 2027, you'll have the right to opt out of algorithms that decide if you get hired, get a loan, or get insurance. The era of the unaccountable "black box" decision is ending, but companies have a year to prepare their excuses.
Your Boss Is an Algorithm
You didn't get the job interview. Your loan application was denied instantly. Your insurance premium went up for no reason.
Why? " The computer said so."
For years, that was the end of the conversation. Companies hid behind "proprietary algorithms" to make life-altering decisions without accountability. They used resume scanners to filter applicants, behavioral profiling to price insurance, and AI surveillance to monitor worker productivity.
California just threw a wrench in the machine.
New regulations for Automated Decision-Making Technology (ADMT) are now in effect. While the full suite of consumer rights kicks in next year, the clock started ticking on January 1, 2026. Companies using these tools must now conduct rigorous risk assessments. They have to prove they know what their monsters are doing.
What Counts as a "Robot Boss"?
The California Privacy Protection Agency (CPPA) defines ADMT broadly. It's not just "AI." It's any system that processes personal information and uses computation to replace (or substantially replace) human decision-making.
If a human just rubber-stamps what the computer suggests? That counts.
The rules cover "significant decisions" that impact your life:
- Employment: Hiring, firing, promotion, discipline, task allocation.
- Finance: Loans, credit limits, interest rates.
- Housing: Tenant screening, lease terms.
- Education: Admissions, financial aid.
- Healthcare: Access to services, insurance coverage.
- Insurance: Premiums, claim denials.
If a machine is deciding your fate in these areas, it's regulated.
The Compliance Ticking Clock
Here's the timeline companies are scrambling to meet:
- January 1, 2026 (Now): Risk assessment requirements begin. Companies must evaluate their ADMT for bias, discrimination, and privacy harm. If they're using a tool that rejects women more often than men, they need to document it now.
- January 1, 2027: The rights go live. Consumers gets the right to opt out of ADMT for significant decisions. You can demand a human review. You also get the right to access information about the logic: companies must explain why the machine said no.
- April 1, 2028: First independent cybersecurity audits are due.
The delay in consumer rights is frustrating, but the risk assessment phase is critical. It forces companies to look under the hood. Many will find their algorithms are illegal under existing anti-discrimination laws.
They Are Watching You Work
The regulations also target "profiling": automated processing that evaluates your personal aspects. Intelligence. Aptitude. Reliability. Behavior. Location.
This hits workplace surveillance hard. Those tools that track your mouse movements to measure "productivity"? The cameras that watch your eyes to see if you're paying attention? That's profiling.
Under the new rules, employees will eventually have the right to know how these systems work and, in many cases, to say no to them.
What You Can Do
Ask the Question Now
If you're denied a service or a job, ask: "Was this decision made by an automated system?" Get it in writing. Even before the 2027 opt-out, companies are nervous. They might escalate you to a human just to avoid scrutiny.
Spot the "Human in the Loop" Fake
Companies will try to dodge these rules by claiming a human reviews the decision. Ask: "Does the human reviewer have the authority to overturn the machine? Do they check the underlying data?" If the answer is vague, it's ADMT.
California Employees: Watch Your Handbooks
Employers are updating privacy notices right now. Look for new sections on "Automated Decision-Making" or "Algorithmic Management." Save these documents. They are your evidence.