If your company does not use AI to manage employees, you may assume California’s new “No Robo Bosses Act” does not apply to you. You may be wrong. SB 947, effective July 1, 2027, reaches automated decision systems (ADS) built on statistical modeling and data analytics, not just AI. It does not ban these systems in discipline and termination, but it requires human corroboration and disclosure, and once ADS use is shown, the employer must prove it followed the rules.
What the New Law Requires
Under new Labor Code section 1522(b), an employer may not “rely solely” on an ADS to discipline or terminate. If it “primarily relies” on an ADS result, what the statute calls an “ADS output," such as a score, flag, or recommendation, a human must corroborate the decision using the underlying data or other supporting information (as supervisor evaluations, personnel files, work product, peer reviews, or witness interviews).
Corroboration is not a rubber stamp. If the result cannot be corroborated, or the reviewer finds it “inaccurate, incomplete, or misleading,” the employer cannot use it (§ 1522(c)). A reviewer who finds no support has not cleared the decision; the reviewer has disqualified the result.
When an employer has primarily used an ADS to discipline or terminate, the employee may request, and the employer must provide, a “meaningful, objective description” of the employee's own data used by the ADS (§ 1522(d)). The description must anonymize the personal information of customers, other employees, and other individuals (§ 1522(e)). Where an ADS relies on coworker comparisons or customer complaints, employers will need to balance useful context against that anonymity requirement.
An employer that primarily relied on an ADS must give the employee a written post-use notice when it delivers the decision (§ 1524(a)). The notice must be written in plain language as a separate, stand-alone communication, in the language used for routine employee communications, and delivered through a simple method such as email or a hyperlink.
The notice must state that (1) the employer primarily relied on an ADS; (2) a human reviewed the decision and corroborated the ADS result; (3) a human contact is available for information about the decision and the right to a description of the employee’s ADS data; and (4) the employer may not retaliate against the employee for exercising these rights (§ 1524(b)).
What It Means for Employers
Five practical points stand out. First, coverage is broad. Section 1520(c) reaches any process derived from “machine learning, statistical modeling, data analytics, or artificial intelligence” that issues a score, classification, or recommendation used to assist or replace human decision making. Basic tools such as spam filters, firewalls, calculators, and databases are excluded, but productivity dashboards, quality scores, attendance algorithms, and monitoring flags may all qualify.
Second, “human in the loop” is not enough. A manager who clicks approve on a flag has corroborated nothing. The reviewer must check the underlying data or independent sources (the personnel file, work product, a supervisor's evaluation, or a witness) and be prepared to reject the result.
Third, the statute does not expressly require written or contemporaneous corroboration, but we recommend both. If a claim surfaces years later, memories will have faded and reviewers may have left; a record made at the time of who reviewed what, and what they concluded, may be the employer’s best evidence.
Fourth, compliance may depend on your vendors. An employer that cannot obtain underlying data from its vendor can neither corroborate the result nor provide the “meaningful, objective description” an employee may request. Solve that in contracting before 2027, not during a dispute.
Fifth, the burden of proof flips. Ordinarily, the party bringing the claim carries the initial burden. Under section 1526.1(c), in a case alleging a violation of section 1522(b) or (c), once ADS use is shown, the employer must prove either that it did not primarily rely on the ADS or that it complied with the corroboration and notice requirements. Both showings depend on documents: what else drove the decision, or who corroborated, what they reviewed, and that a complete notice was sent. Testimony that a manager “would have” decided the same way will not suffice.
The bottom line: an employer with a thin file can lose a case it should have won, even if the decision itself was fully justified. And the stakes are real. The Labor Commissioner and public prosecutors may enforce the law, and remedies include injunctive relief, punitive damages, attorney’s fees, and a $500 civil penalty per violation. Good recordkeeping is the defense.
What We Don't Know Yet
No regulations or case law interpret the statute yet, leaving several open questions.
- What “primarily relies” means. The phrase triggers corroboration, data-access, and notice duties but is undefined. Until guidance arrives, a broad reading is safer as an employer that guesses wrong will have skipped corroboration and notice and will bear the burden of proving otherwise.
- What showing establishes that an ADS was “used.” Burden shifting turns on this showing, but how specific the proof must be and whether a scoring tool in the workflow suffices is unresolved.
- How much corroboration is enough. The statute lists categories of supporting information but does not say how many sources or how deep the review must be.
- Which tools are covered. The definition discussed above leaves its edges unclear: when a simple rule-based scoring or flagging tool counts as “data analytics,” and when a result “materially impacts” an employee rather than merely informing a manager.
- Whether employees can sue directly. The bill names the Labor Commissioner and public prosecutors as enforcers but does not expressly create a private right of action, leaving open whether employees can sue on their own or under PAGA.
The Takeaway for Employers
- Identify tools whose scores, rankings, or flags could feed a discipline or termination decision, including tools adopted by business units without IT or HR involvement.
- Treat any decision meaningfully driven by an ADS result as covered, rather than litigating internally over whether reliance was “primary.”
- Build a corroboration record into the disciplinary workflow, captured at the time of the decision: who reviewed, what they consulted, and whether they confirmed or rejected the result.
- Give reviewers explicit authority, in writing, to reject an ADS result, and make clear that an uncorroborated result cannot support discipline.
- Draft a post-use notice template now that covers all four required elements and meets the format rules.
- Set up a process for employee data requests, including who responds, the format of the description, and how to anonymize third-party and peer-comparison data.
- Train managers and HR on what corroboration requires and on the prohibition against retaliation for exercising these rights.
- Review and, where necessary, renegotiate vendor contracts to secure access to underlying data, explainability support, and cooperation in litigation.
Compliance takes time. Employers that build corroboration and documentation into their workflows now, and test them before July 1, 2027, will be far better positioned when the first claims arrive. If you have any questions or need specific guidance on how these changes may impact your organization, please email Sweta Patel directly (spatel@potomaclaw.com) to discuss further. If you would like to read more on Sweta's Employment updates, please subscribe to her list here.
This article is for informational purposes only, is not legal advice, and does not create an attorney-client relationship. The statute has not yet taken effect, and no regulations or judicial decisions interpret it. Employers should consult counsel about their specific circumstances. Please contact Sweta H. Patel at Potomac Law Group with any questions.
