The “No Robo Bosses” Act is no longer a future compliance problem.
California SB 7 took effect January 1, 2026. It’s April. Employment law practitioners report that compliance reviews are now underway at California employers. The law hasn’t generated its first publicized enforcement action yet, but that’s a timing question, not a certainty question. The compliance posture that matters right now is not reactive. It’s whether your organization can produce documentation on demand.
This piece maps the five operational compliance questions that mid-market California employers need to answer. These are the questions an auditor or plaintiff’s attorney would ask. If your answer to any of them is uncertain, that’s where the work is.
Does This Law Apply to Your Organization?
California SB 7 applies to employers using Automated Decision Systems in covered workforce decisions. The employer headcount threshold is an important scoping question that requires confirmation against the full bill text, secondary summaries vary in their characterization of the threshold, and the Labor Code is the authoritative source. Any employer using AI-assisted tools for hiring, performance management, scheduling, or termination decisions in California should assume applicability until legal counsel confirms otherwise.
What counts as an Automated Decision System? The California Labor Code’s statutory definition is the governing text. In general, ADS covers computational tools that use algorithms, machine learning, or AI to inform or make decisions about individuals. If your HR software uses AI-generated recommendations, even recommendations that a human nominally reviews – that system likely qualifies.
The “sole factor” prohibition is the most clearly defined requirement. AI cannot be the only input to a termination or discipline decision. Human judgment must be demonstrably part of the process. What “demonstrably” means in an audit context is the operational challenge.
What Goes in an ADS Registry?
The mandatory ADS registry is the foundation of SB 7 compliance. You cannot demonstrate human oversight for systems you haven’t enumerated. The registry requirement is both a compliance artifact and a compliance process, building it forces the internal audit that identifies which systems are actually in scope.
SHRM’s analysis of SB 7 notes that many mid-market employers have accumulated AI-assisted HR tools incrementally, an ATS with AI screening, a performance management platform with algorithmic flagging, a scheduling tool with predictive optimization, without treating them as a unified system category. The registry requirement makes that incremental accumulation visible and auditable.
At minimum, the registry should capture: the system name and vendor, the workforce decision domain it affects (hiring, scheduling, performance, termination), the decision type it informs (recommendation, flag, automated action), and the human review process associated with each. The Labor Code’s specific registry format requirements should be confirmed against the statutory text, what we know from secondary sources is that the requirement exists; what it specifically mandates belongs in the bill’s language, not a summary.
One practical reality: vendors may not characterize their products as ADS. An applicant tracking system that “helps identify strong candidates” and a performance management tool that “surfaces performance trends” are both potentially in scope. The employer’s registry obligation runs to what the systems do, not what the vendors call them.
How Do You Document Human Review?
The human-in-the-loop requirement is only as strong as its documentation. A human who nominally reviews an AI recommendation but has no practical ability to override it, no documented basis for their review, and no training on what the review is supposed to accomplish is not a meaningful human in the loop. An audit would challenge that review process directly.
Defensible human-review documentation has three components. First, the reviewer must have access to the information they need to exercise judgment, not just the AI’s output, but the underlying data or employee record the AI evaluated. Second, the review must be recorded: who reviewed, when, what information they considered, and what decision they reached. Third, deviation from the AI’s output must be possible and must have happened in at least some cases, a review process that always confirms the AI’s recommendation is substantively indistinguishable from using AI as the sole factor.
The law’s human-oversight requirement applies to “high-stakes” workforce decisions. Termination is the enumerated example. Whether performance improvement plans, demotions, or scheduling changes that effectively reduce earnings qualify as high-stakes decisions is an interpretive question that the first enforcement action or litigation will clarify. The conservative compliance posture is to build human-review documentation for any decision where AI influence is significant and where the outcome materially affects an employee’s status or compensation.
What Does an SB 7 Audit Actually Examine?
No enforcement action has been publicly reported as of publication. That means the audit framework is still being defined by the statutory text and the practical expectations of employment law practitioners, not by documented enforcement precedent.
Based on the law’s requirements and California labor enforcement patterns, an audit would likely examine: the completeness of the ADS registry (are all systems captured?), the adequacy of human-review processes for covered decisions (is review genuine, documented, and meaningful?), the training provided to human reviewers (do they understand what the AI is doing and what their review is supposed to accomplish?), and notice to employees (does the law require disclosure to affected employees about ADS use in decisions about them?, confirm against bill text).
The first publicized enforcement action will define what “compliance” looks like in practice more concretely than any legal analysis written before that action. Employment law counsel who specialize in California labor law are the right resource for specific audit preparation, not generic AI compliance guidance.
What Are Peer Jurisdictions Doing, and How Does SB 7 Fit?
SB 7 is California’s most direct expression of the human-oversight principle in employment contexts, but it’s not operating in isolation.
The EU AI Act’s workplace AI provisions, which classify certain AI systems used in employment contexts as high-risk, require conformity assessments, technical documentation, and human oversight as mandatory compliance elements for EU market deployment. The EU’s approach to certifying AI systems in high-stakes contexts provides a useful reference framework even for US-only employers: the documentation standards the EU requires for high-risk employment AI are a reasonable benchmark for the documentation any employer should be building under SB 7.
The federal vs. state AI regulatory tension is directly relevant here. Any federal AI preemption legislation that excluded employment law would leave SB 7 standing. California’s legislative history on labor matters suggests the state would resist preemption of employment-specific AI protections specifically. Employers who treat SB 7 as a temporary compliance burden pending federal resolution are making a planning assumption that may not hold.
The Connecticut Senate’s April 22 passage of an AI safety bill, which addresses different AI harms (content moderation, chatbot safety) but reflects the same pattern of state-level action filling federal gaps, confirms that the multi-state AI compliance reality is not a California-specific situation. The patchwork is building.
The Bottom Line
Four months into SB 7’s effective date, the compliance question is operational, not strategic. The strategy was settled when the law passed: if you’re a California employer using AI in workforce decisions, you comply. The operational questions, what goes in the registry, how you document human review, which decisions trigger the requirement, are answerable now, before an enforcement action forces the answer under adversarial conditions.
Two documents drive this. The full California Labor Code text for SB 7 is the authoritative source for every specific requirement. Employment law counsel with California labor specialization is the right resource for translating that text into a defensible compliance program. Secondary summaries, including this one, are useful for orientation. They’re not a substitute for either.