California's SB 1162 first covered frontier AI labs in the 2026 filing cycle — and the headline figure from that inaugural data is a 22% median earnings gap between men and women in AI engineering roles, wider than any other tech discipline on record. Employers with 100 or more California-based employees, including Anthropic and OpenAI crossing that threshold for the first time, were required to disclose mean and median hourly wages by job category, sex, race, and ethnicity. The California Civil Rights Department accepted inaugural filings through August 2026 for companies that only recently crossed the 100-employee mark.
The CRD's aggregate analysis of 2024 Reporting Year data, published March 26, 2026, covers the Computer and Mathematical Occupations category — the classification that now encompasses AI engineering, machine learning research, and AI safety roles. The gap in broader software development is 14%. The gap across all California private-sector categories is 12%. AI is the outlier, and the data is on file with the state.
What SB 1162 Actually Requires
California's pay data reporting obligation, originally established by SB 973 and substantially expanded by SB 1162 in 2022, requires private employers with 100 or more employees to file annual reports with the CRD detailing their workforce by pay band, job category, sex, race, and ethnicity. The 2025 Reporting Year cycle — covering pay data from calendar year 2025 — carried a primary deadline of May 13, 2026, with first-time filers granted extended submission windows through August 2026.
The 2026 cycle introduced two meaningful changes. First, the CRD is transitioning from 10 legacy EEO-1 job categories to 23 Standard Occupational Classification codes, per the agency's posted transition guidance, with the new framework planned to take effect for reports due in May 2027. For the 2025 RY cycle, 10 categories remain in force, but Category D, Computer and Mathematical Occupations, now acts as the effective container for AI-specific roles. Second, and more consequentially for AI labs specifically, SB 464 — signed by Governor Newsom in October 2025 and effective January 1, 2026 — transformed the penalty structure. Fines are now mandatory, not discretionary. A company that fails to file faces $100 per employee on the first violation, $200 per employee on subsequent failures. A 200-person frontier lab that misses its inaugural filing carries $20,000 in immediate exposure. "Good faith" effort is no longer a statutory defense.
For companies like Anthropic and OpenAI, this reporting cycle carries new weight under SB 464's mandatory penalty regime, effective January 1, 2026: failure to file now triggers automatic civil liability rather than discretionary compliance letters. The numbers they have submitted are not individually identified. But in aggregate, as the CRD publishes its statewide analysis, the sector-level picture is now visible.
What the Filings Reveal: The 22% Gap
The CRD's March 2026 publication of 2024 Reporting Year aggregate data — covering nearly 8 million workers across California's large private-sector employers — showed that women and people of color remain disproportionately concentrated in lower pay bands. For technology roles broadly, the pattern is not new. What the SB 1162 data clarifies, for the first time with mandatory employer-submitted figures, is the magnitude inside AI specifically.
Women hold approximately 22% of AI roles nationally, according to Women in Tech Network data compiled through Q1 2026, and 18% of machine learning engineering positions specifically. That representation deficit translates directly into a pay gap because AI compensation is heavily right-skewed: the largest packages cluster at senior and principal levels, and those levels are disproportionately male. When the CRD computes median hourly rates for Category D (Computer and Mathematical Occupations) by sex across filings from California employers, the resulting 22% gap reflects not a like-for-like pay disparity within identical roles but a structural concentration of women at lower bands within the same broad job category.
The distinction matters for employers — and for the enforcement direction the CRD is signaling. A company that pays its male and female senior ML engineers the same base salary but has three times as many men at the senior level as at the associate level will show a 22% gap in the CRD's median comparison. That is a workforce composition problem, not a direct pay discrimination problem. But it is the same gap. And it is now documented.
For AI labs specifically, several additional dynamics compound the gap:
Equity concentration. The largest compensation differentials in AI are in equity, not base salary. A senior male researcher at Anthropic or OpenAI hired pre-2024, when valuations were lower and options were struck at earlier prices, holds substantially more unrealized equity than a female associate engineer hired in 2025 at a higher strike price. SB 1162 captures mean and median hourly rates, not equity value. The reported gap understates the realized compensation divergence.
Band width opacity. As covered in ENTRA's August 1 briefing on the three-state salary transparency framework, posted bands at AI labs span 50% to 67% of the floor. That spread is wide enough to accommodate two distinct compensation tiers within a single legal level. SB 1162 requires disclosure of what employers actually paid by band; it does not require disclosure of where within a band each demographic group clusters. Until the 2027 transition to 23 SOC categories, which will allow the CRD to compute median pay for AI-specific sub-classifications rather than the broad Computer and Mathematical bucket, that granularity remains unavailable.
Company-by-Company: Who Filed What
California does not publish individual employer SB 1162 filings; it publishes aggregated sector-level analysis. What individual companies disclose beyond the statutory minimum is a matter of voluntary transparency. Here is where the six companies in scope for this briefing stand.
Google. Google's California engineering workforce runs into the tens of thousands and spans multiple campuses — Mountain View, Sunnyvale, and the growing San Francisco AI hub. Google has published voluntary pay equity analyses since 2018 and settled a $118 million class-action gender discrimination suit in 2022, with the settlement covering more than 15,000 female employees. Its internal pay equity analysis, disclosed in annual diversity reports, claims to find no statistically significant pay gap on a role-adjusted basis. What its SB 1162 filing shows — before role adjustment — is a different picture, one the CRD's aggregate analysis reflects in the 22% figure for Computer and Mathematical Occupations. Google's 2024 Diversity Annual Report, released in 2025, showed 35% female representation in its global technical workforce, a number that has moved less than two percentage points over five years.
Apple. Apple's Cupertino campus is California's largest single-employer technology site. Apple's AI engineering hiring, concentrated in its Vision Products Group, Core ML teams, and the Apple Intelligence engineering organization formed after WWDC 2024, has grown substantially in 2025 and 2026. Apple publishes EEO-1 data voluntarily and has disclosed pay equity analyses claiming less than one cent per dollar gap on a controlled basis. Its SB 1162 filing for 2025 RY will be the most granular pay data Apple has ever submitted to a regulator, covering median hourly rates by job category and sex.
Meta. Meta's California headcount has contracted significantly through its 2025 and 2026 restructuring rounds: approximately 600 positions eliminated in Meta's AI division in October 2025, followed by 8,000 company-wide in May 2026. Meta Superintelligence Labs, now running under Chief AI Officer Alexandr Wang, is the primary remaining California AI engineering organization. Representation data from Meta's most recent diversity report shows female employees comprise 24% of its technical workforce globally. The SB 1162 filing for the 2025 reporting year covers the period of maximum flux in Meta's California AI headcount, making the median hourly rate calculation for Computer and Mathematical Occupations a snapshot of a workforce mid-restructuring rather than a stable baseline.
Anthropic. Anthropic's San Francisco headquarters has been subject to SB 1162 pay data reporting since 2023, when its California headcount crossed the 100-employee threshold. The 2025 Reporting Year cycle is the company's first filing under the SB 464 mandatory-penalty regime — automatic civil liability, not discretionary CRD compliance letters, now governs non-compliance. Anthropic has not published a standalone DEI or pay equity report. Its annual California filings represent the only externally mandated disclosure of gender pay distribution data at the company. Levels.fyi data tracking Anthropic compensation through Q2 2026 shows a company-wide median total compensation of approximately $420,000, with significant right-skew toward principal and distinguished researcher levels. What the SB 1162 filing adds that Levels.fyi cannot: the demographic breakdown of who sits at those levels.
OpenAI. OpenAI's San Francisco headquarters places it squarely within California's jurisdiction. Its workforce has grown from approximately 3,000 employees at end of 2024 to a projected 8,000 by December 2026. OpenAI published a 2024 DEI report disclosing that 16% of its technical roles were held by women at the time of publication. Under SB 1162, that 16% representation in technical roles, combined with compensation concentration at senior levels, produces the median hourly gap the CRD is now able to calculate and publish in its aggregate analysis. OpenAI's compensation structure — base plus Profit Participation Units priced at valuation — makes the base-only median in the SB 1162 filing a partial picture, but it is a picture that now exists in the public record.
Salesforce. Salesforce is the outlier in this peer set, and worth including for exactly that reason. The company has conducted annual pay equity analyses since 2015, spent more than $22 million in cumulative pay adjustments to close identified gaps, and has published detailed methodology on its approach. Its SB 1162 filings are not its first regulated pay disclosure. Salesforce's most recent public equal pay assessment, covering 2025 data, reports that it found and closed gaps affecting 6% of its global employee population. Its California filing for 2025 RY will be the most scrutinized validation of that claim — the difference between a self-reported pay equity analysis and a regulator-mandated pay data submission is the absence of self-selection in what gets included.
What Comes Next
Three enforcement and legislative developments will shape what happens to this data between now and the next filing cycle.
CRD enforcement posture under SB 464. The shift from discretionary to mandatory penalties, effective January 1, 2026, has already altered compliance behavior. The CRD had issued compliance letters to non-filers in prior cycles; in 2026 it is authorized to go directly to civil action. For AI labs that filed inaugural reports in this cycle, the next step is verification: the CRD cross-references submitted pay data against EDD wage records for California employees. A filing that underrepresents median pay or misclassifies roles into lower job categories is identifiable. The agency's enforcement posture after EPEWA-style enforcement ratcheting in Colorado — where the CDLE issued more than 600 compliance letters and pursued active civil enforcement through June 2026 on band-width and omission issues — suggests California's CRD is watching Colorado's trajectory and drawing operational conclusions.
The 2027 transition to 23 job categories. When the SOC-based 23-category framework takes effect for reports due May 2027, AI-specific roles will no longer be bundled into the broad Computer and Mathematical Occupations category. Machine learning research, AI safety, and AI engineering will appear as distinct classification groups. The CRD will be able to compute and publish median pay gaps for each. That granularity will expose what the current 10-category framework obscures: whether the 22% gap is driven by the highest-paying AI sub-roles being male-concentrated, or whether a within-role gap exists that the category-level aggregate conceals. For AI labs, 2027 is the year when the SB 1162 data becomes genuinely company-comparable at the discipline level.
Will California become the de facto national standard? The trajectory of state-level pay data legislation runs in one direction. Colorado's EPEWA, effective since 2021, produced more than 2,800 complaints and more than 600 compliance letters in five years of operation. California's SB 1162, covering a workforce roughly six times larger than Colorado's, is in year three of operation with mandatory penalties now in force. Illinois, Washington, Massachusetts, and New Jersey have advanced salary disclosure legislation in 2025 and early 2026. Industry estimates through Q1 2026 project that by January 2027, more than 65% of the US workforce will operate under some form of pay transparency requirement, as state-level mandates continue expanding beyond California, Colorado, New York, and Illinois.
The practical effect for AI labs is that California's reporting framework — which requires not just pay scale disclosure on job postings but actual median pay data by demographic group — will become the baseline expectation nationally before 2028. Companies that have invested in pay equity infrastructure, like Salesforce, are positioned to absorb that requirement without operational disruption. Companies filing mandatory pay data reports for the first time in 2026, including Anthropic and OpenAI, are in the early-stage compliance posture: meeting the letter of the law, without yet having built the analytics infrastructure to understand, and act on, what their own data shows.
The 22% gap in AI engineering is not new. It is newly documented. That is what changes in August 2026.
Methodology note: The 22% median earnings gap for Computer and Mathematical Occupations cited in this briefing is drawn from the California Civil Rights Department's aggregate analysis of 2024 Reporting Year pay data, published March 26, 2026. Individual company filings are not publicly identified by the CRD; aggregate data is published by sector, job category, pay band, and demographic group. Company-level workforce representation figures are drawn from voluntarily published diversity reports and, where noted, Levels.fyi self-reported compensation data through Q2 2026. Pay data from SB 1162 filings captures mean and median hourly rates; it does not capture equity compensation, which constitutes the majority of total compensation at frontier AI labs. Reported figures should be treated as lower-bound estimates of the actual compensation gap.
Global Career PlatformFind AI talent. Find your next role.
Booking is hotels. · Airbnb is apartments. · ENTRA is global careers.
Sources: California Civil Rights Department — March 2026 pay data release — SB 464 mandatory penalties guide, JDSupra/Amundsen Davis — California SB 464 enforcement, National Law Review — California pay data reporting 2026 requirements, Stoel Rives — SB 464 tougher pay reporting, Shaw Law Group — Women in AI statistics 2026, Women in Tech Network — Gender pay gap in tech 2026, WomenHack — OpenAI DEI report 2024, Makerstations — Google $118M gender pay settlement, UCI — California pay data reporting portal, CRD — 2026 gender pay gap report, Payscale — Salesforce pay equity, Salesforce.com