Forty-eight percent of active AI job postings carried some form of salary disclosure in Q2 2026; yet the skills commanding the largest compensation premiums in the same period disclose at rates as low as 19 percent. That structural inversion, documented across 47,200 postings in the ENTRA Q2 2026 Job Signal Index and confirmed by the ENTRA Q2 2026 Salary Survey across 2,140 practitioners in 47 countries, is the defining pattern of Salary Transparency Month: the information gap in AI hiring is closing fastest where it matters least, and holding open where it matters most.[^1]
The regulatory infrastructure that forced disclosure from voluntary to mandatory is no longer contained to one jurisdiction. Six US states now require salary ranges in postings. The EU Pay Transparency Directive passed its June 7, 2026 transposition deadline. The UK Employment Rights Act 2025 obligates pay gap action plans by April 2027. From Q4 2026, a company posting AI roles across San Francisco, London, and Bratislava simultaneously faces three enforcement frameworks with distinct mechanics, different penalty thresholds, and no coordination between them. What happens to AI compensation architecture when every major employer jurisdiction moves toward mandatory disclosure at roughly the same time is the operational question this month's data have been building toward.
This report is the synthesis. It draws on four ENTRA datasets (the Pay Transparency Audit, the Salary Survey, the Job Signal Index, and CHRO Survey fieldwork) to map what employers actually did when disclosure became non-optional, what candidates did with the data when they got it, and where the regulatory arc points in Q4 2026 and 2027.
Section 1: The employer response
When disclosure became mandatory or competitively expected, the 87 companies in the ENTRA Pay Transparency Audit Q2 2026 sorted into three identifiable camps. The camps do not map cleanly onto company size or funding stage. They map onto a single internal decision: whether compensation is a candidate education tool or a competitive intelligence shield.[^2]
| Camp | Share of audit cohort | Avg. audit score | Primary behavior | |---|---|---|---| | Disclosers | 24% (21 companies) | 78/100 | Full band, posted globally; equity structure included | | Compliers | 47% (41 companies) | 58/100 | Band in mandate states; spread ≥$150K (informationally minimal) | | Resisters | 29% (25 companies) | 31/100 | No bands; withdrawn state postings or agency-routed hiring |
Source: ENTRA Pay Transparency Audit Q2 2026 (n=87 companies). Camp thresholds: Discloser ≥70/100, Complier 40–69/100, Resister below 40/100.
Disclosers
Hugging Face (85/100, AA), Stripe (82/100, AA), and GitHub (75/100, A) define the Discloser camp. All three post salary bands globally, not solely in mandate jurisdictions, and include equity structure alongside base ranges. Hugging Face links its compensation benchmarking methodology directly from role descriptions. Stripe posts salary bands in the United Kingdom and France ahead of any legal obligation to do so. GitHub's Copilot engineering postings carry the full Microsoft SB 1162-compliant band plus a stated equity grant range.
The funnel consequences of that posture are measurable. Discloser-camp companies in the ENTRA audit reported application volumes 34 percent higher per open role than Complier-camp companies, with recruiter screen-to-interview conversion rates of 68 percent against 49 percent. Offer acceptance rates at Disclosers averaged 81 percent; at Compliers, 64 percent. That 17-percentage-point gap translates (at 50 senior hires annually and an average sunk recruiting cost of $68,000 per declined offer) to approximately $578,000 in wasted recruiter investment per year, before accounting for time-to-fill delay.[^3]
Time-to-hire compounds the arithmetic. Discloser-camp companies filled equivalent senior AI roles in an average of 46 days. Complier-camp companies: 55 days. Resister-camp: 68 days. The BLS JOLTS June 2026 data placed AI-adjacent tech job openings at 847,000, a 14-year high. In a market that tight, 22 additional days to fill each role is not a recruiting inconvenience. It is a capital allocation decision.[^4]
Compliers
Anthropic (62/100, BBB) and OpenAI (65/100, BBB) represent modal frontier-lab behavior under US state disclosure mandates. Salary ranges appear in Colorado, New York City, California, and Washington postings and effectively nowhere else. The spreads posted in mandate jurisdictions are wide: Anthropic's L5 ML engineer band runs $120,000 to $320,000, a $200,000 spread that satisfies every applicable state law and tells a senior candidate nothing about where within that range they would actually land.[^5]
The wide-band architecture at frontier labs is not organizational carelessness. It is a deliberate response to the dual-track compensation structure that has become standard across the top AI tier. A research-engineer track ($480,000 to $740,000 total comp at L6 at Anthropic) and a product-engineer track ($360,000 to $540,000) coexist within the same role family on the same external ladder. Collapsing both into one compliant band requires a spread wide enough to hold both tracks without disclosing the differential. The compliance is real. The disclosure is not.[^6]
OpenAI's 65/100 score reflects an additional complication: the 2025 profit-participation-unit-to-RSU equity conversion restructured the equity component of total compensation across a full fiscal year without any candidate-accessible documentation explaining how the converted structure compared to what it replaced. Legally compliant; functionally opaque on the dimension the ENTRA audit weights at 25 of 100 points.[^7]
Resisters
Twenty-five companies in the audit cohort scored below 40/100. Their tactics cluster into three observable behaviors: withdrawing postings from high-mandate states (Colorado, New York), routing senior AI hires through executive search firms whose posting practices fall outside state disclosure mandates, and fragmenting job title strings (substituting "Conversational AI Infrastructure Principal" for "Senior ML Engineer") to evade existing category audits.[^8]
xAI scored 28/100 in the ENTRA audit, the lowest score among frontier-adjacent labs in the cohort. The company posted the majority of its Q2 2026 senior AI roles as US-remote without jurisdiction-specific salary ranges, despite California SB 1162's applicability to any employer with 15 or more employees that posts roles to California candidates.[^9]
The economic logic of resistance is narrowing. Resister-camp companies averaged 68 days to fill senior AI roles, 22 days longer than Disclosers and 13 longer than Compliers. In a 847,000-opening market, the candidate who encounters opacity at a Resister-camp employer has alternatives posted with visible bands at 48 percent of the market. That share has risen each quarter in 2026.
Section 2: The candidate shift
When candidates can see a salary band before the first screen, four measurable behaviors change: negotiation outcomes, application abandonment rates, geographic search patterns, and tolerance for wide-band disclosure theater.
| Posting condition | Avg. year-one uplift | Opening offer (% of band) | Offer-to-close rate | |---|---|---|---| | Tight band visible (spread below $80K) | +$43,000 vs. baseline | 61st percentile | 79% | | Wide band visible (spread above $150K) | +$18,000 vs. baseline | 44th percentile | 68% | | No band posted | Baseline | N/A (peer anchor only) | 61% |
Source: ENTRA Q2 2026 Salary Survey (n=2,140 AI practitioners, 47 countries). Total comp: base + annualized equity at grant price + cash bonus.[^10]
The $43,000 negotiation uplift documented in the August 7 pay transparency dividend report and the August 21 regulatory endgame analysis is the most cited finding of Salary Transparency Month. But the table above adds a finding the earlier reports previewed without fully quantifying: wide-band compliance theater delivers roughly 40 percent of the candidate uplift that meaningful disclosure delivers ($18,000 against $43,000). A Complier-camp band moves the needle. It does not reset it.
Job abandonment. When the posted band falls below a candidate's stated compensation floor, 63 percent of ENTRA survey respondents did not submit an application. Three percent applied regardless. Thirty-four percent contacted the recruiter to clarify before deciding.[^11]
That 63 percent abandonment rate explains the strategic function of wide-band compliance. A $120,000-to-$320,000 band does not trigger abandonment because the $320,000 ceiling clears most senior candidates' floor. A tighter, more honest band (say, $120,000 to $175,000 for a role that actually pays in that range) triggers abandonment from candidates whose market rate is $220,000. Complier-camp employers use band width to suppress the abandonment signal. Candidates with peer networks sophisticated enough to cross-reference Levels.fyi Q1–Q2 2026 data, ENTRA benchmarks, and recruiter conversations see through it. Candidates without those networks do not.
The transparency paradox: highest premium, lowest disclosure. The skills commanding the largest compensation premiums in the ENTRA Q2 2026 Job Signal Index disclose at the lowest rates.
| Skill | Pay premium vs. median ML | Posting disclosure rate | |---|---|---| | Constitutional AI | +34% | 19% | | RLHF | +28% | 31% | | Multimodal systems | +22% | 38% | | LLM inference optimization | +18% | 44% | | Python / PyTorch (baseline) | 0% | 71% |
Source: ENTRA Q2 2026 Job Signal Index (n=47,200 postings). Disclosure rate = share of postings requiring the skill that include a salary band. Pay premium vs. median Senior ML Engineer.[^12]
Constitutional AI safety commands a 34 percent pay premium against the median ML engineer baseline. Its disclosure rate is 19 percent, the lowest in the ENTRA skill taxonomy. RLHF carries a 28 percent premium and a 31 percent disclosure rate. Section 4 of this report maps the four structural mechanisms that produce this inversion.[^13]
Geographic arbitrage. Thirty-seven percent of US-based ENTRA survey respondents located outside Colorado, California, New York, Washington, New Jersey, and Illinois reported filtering for roles in mandate states specifically to access posted salary bands, including when the role was listed as US-remote.[^14]
The implication for Resister-camp employers is structural. Posting roles as US-remote without bands works as a transparency avoidance mechanism only until candidates realize the same role, posted under a Colorado or California location slug, carries a legally mandated band. For companies with genuinely national hiring pools, the band is de facto public whether or not they post it nationally.
Section 3: The global regulatory arc
| Framework | Status (Aug 2026) | Q4 2026–2027 milestone | |---|---|---| | US states (CO, CA, NY, WA, NJ, IL) | Active enforcement; NJ SB1568 effective June 2026 | IL threshold drops to 100+ employees Jan 2027 | | US federal PALS Act | Proposed; Senate vote pending | Senate Commerce Committee markup Q4 2026 | | EU Directive 2023/970/EU | 23 of 27 missed June 7 deadline; 4 transposed | Enforcement begins in Slovakia, Estonia, Lithuania, Malta; EC Article 258 proceedings | | UK Employment Rights Act 2025 | Royal Assent Dec 18, 2025 | Action plans due April 2027 (250+ employees) | | Saudi MOHRE Resolution 0340 | 12-month implementation window | Q1 2027 compliance review |
EU enforcement. Twenty-three of 27 EU member states missed the June 7, 2026 transposition deadline for Directive 2023/970/EU. The four that completed transposition on time (Slovakia, Estonia, Lithuania, and Malta) are the first jurisdictions where Directive enforcement can proceed against individual employers. Member states that transposed the Directive have established penalty regimes; enforcement frameworks vary by jurisdiction, with some targeting a percentage of annual payroll.[^15]
The European Commission opened Article 258 infringement proceedings against non-transposing member states in July 2026. Under standard CJEU sequencing, referrals from the Commission's reasoned opinions, expected to issue in Q4 2026 for the most recalcitrant non-transposers, can reach the Court within 60 to 90 days. For AI employers in non-transposing states, the operative question is whether national labor inspectorates will enforce the Directive through direct effect doctrine ahead of implementing legislation. The first enforcement action against a named employer in a transposing jurisdiction (expected in Slovakia or Estonia) will set the market's estimate of enforcement risk across all 27 states.[^16]
US federal PALS Act. The Pay Equity for Access to Lucrative Secrets Act, as proposed, would extend salary disclosure obligations to all federal contractors and employers with 100 or more employees nationwide. As of August 28, 2026, the Act is proposed legislation with a Senate vote pending. The Senate Commerce Committee has scheduled markup for Q4 2026. If enacted in its current form, the PALS Act closes the most consequential gap in the US disclosure landscape: the AI employer with 30 employees in a non-mandate state that faces zero current disclosure obligation for a senior research role paying $600,000 total comp. ENTRA's Q2 2026 CHRO Survey found 64 percent of AI-sector CHROs view the PALS Act as likely to pass in some form by mid-2027, a figure that implies most major employers are already modeling compliance costs.[^17]
Illinois. The Illinois Equal Pay Act extension, effective January 1, 2027, lowers the employee threshold for pay data reporting from 250 to 100. Illinois hosts the third-largest AI employer density by headcount after California and New York, concentrated in Chicago's financial district, including JPMorgan AI Research, Citadel's quantitative ML teams, and a cluster of applied AI functions at major banks. The threshold change brings full reporting obligations to that tier for the first time.[^18]
New Jersey. New Jersey SB1568, effective June 2026, requires salary range disclosure in all job postings for employers with five or more employees in the state. The five-employee threshold is the most permissive in the US landscape, covering effectively every meaningful AI employer with New Jersey operations, including NVIDIA's New Jersey-proximate data center operations and Morgan Stanley's AI research functions in the New York metro.[^19]
UK ERA 2025. The Employment Rights Act 2025 moves the UK disclosure standard from reporting to accountability. Prior gender pay gap reporting required a number. Action plans, required from employers with 250 or more employees by April 2027, require a named responsible officer, a documented remediation plan, and published milestones. For AI employers like Microsoft (Reading and London), Google DeepMind (London HQ), and Meta AI (London office), the action plan creates a public accountability record that exists under no current UK or EU framework. The ACAS/EHRC guidance on whether equity-weighted total compensation counts toward the pay gap calculation remains unresolved as of August 2026. If equity is included, disclosed gaps at frontier-lab-affiliated UK employers (where equity comprises 40 to 60 percent of total comp at the senior research level) will be materially wider than base-salary-only figures suggest.[^20]
Saudi Arabia. Vision 2030 Saudization targets create implicit pay band pressure for international AI hires through MOHRE Resolution 0340. The resolution's 12-month implementation window closes in Q1 2027, requiring Saudization-compliant employers to document compensation bands for both Saudi and international staff. International AI employers with Saudi entity structures, including Microsoft Azure Saudi Arabia, Google Cloud Riyadh, and NVIDIA's Gulf commercial operations, face band posting requirements with no parallel in any other regional framework.[^21]
ENTRA's forward estimate: by January 2027, 74 percent of Fortune 500 AI postings globally will carry some form of salary disclosure, up from the current 48 percent. The gap between those two numbers is where Q3 and Q4 2026 employer decisions will play out.[^22]
Section 4: The transparency paradox explained
The central editorial finding of August 2026 is not that disclosure is spreading. It is that the skills, companies, and geographies where transparency would do the most to correct information asymmetry are precisely where disclosure rates are lowest. Constitutional AI at 19 percent. RLHF at 31 percent. The two skills commanding the largest pay premiums sit below the 35th percentile of disclosure rates across all tracked AI skills. Four structural mechanisms produce that inversion.
Mechanism 1: Competitor intelligence. A frontier lab that publishes a meaningful band for Constitutional AI safety research roles is publishing a pricing signal to every competing employer recruiting from the same 300-to-400-person global talent pool. When Anthropic posts a Research Scientist band of $480,000 to $740,000 total comp at L6, OpenAI's recruiting team reads that number before the first candidate does. For skills where the addressable talent pool is this thin, employers treat the posted band as a competitive intelligence document. Non-disclosure is a trade secret posture applied to compensation, and it operates with the same logic.[^23]
Mechanism 2: Incumbent retention risk. An existing employee who sees a posted band for their own role and finds themselves at the 28th percentile of a range extending to $740,000 has a documented basis for a compensation review conversation. In aggregate, when employers post honest bands for high-premium skills, the internal equity review load increases immediately. ENTRA's Q2 2026 CHRO Survey found that 71 percent of CHROs at AI employers who began posting meaningful bands reported a measurable increase in internal compensation review requests within 90 days of the first posting cycle. The majority of those requests came from employees at or below the band midpoint, the cohort concentrated at the floor by the historical inequities documented in the August 14 equity analysis.[^24]
Mechanism 3: Equity exposure. The California Civil Rights Department 2025 Pay Data Report, the first mandatory cycle under SB 1162's reporting provisions (published May 2026), found a mean unadjusted gender pay gap of 17.4 percent in professional computing occupations across technology-sector employers. Within that aggregate, high-premium skill categories showed above-average between-employee variance: Constitutional AI safety roles, ML infrastructure at the senior tier, and RLHF fine-tuning positions all exceeded the sector mean. Employers with high within-band variance know that posting a meaningful band makes the distribution observable to candidates, regulators, and existing employees simultaneously. A $120,000-to-$320,000 spread contains the variance. A $120,000-to-$175,000 spread exposes it.[^25]
Mechanism 4: Supply scarcity. The fundamental mechanism is supply-and-demand simple: employers disclose compensation when they need to attract candidates who have alternatives. When the talent pool for a given skill is small enough that every qualified candidate is already known to every employer recruiting for it, the band is not a recruiting tool. It is a compliance artifact. With an estimated 300 to 400 globally qualified Constitutional AI safety researchers, frontier labs recruiting from that pool are not broadcasting to a job board. They are conducting outreach to named individuals. Those individuals do not need a posted band to initiate a conversation; the employer's offer itself serves as the anchor at the moment of negotiation. Scarcity makes voluntary transparency unnecessary. Regulation is the only mechanism that makes it mandatory regardless of market conditions.
These four mechanisms operate simultaneously and reinforce each other. The result is a market where the highest-premium skills are also the most informationally asymmetric, and the candidates who could most benefit from disclosed bands are those with enough seniority and peer network access to extract the information anyway.
Methodology
This report synthesizes four primary data sources.
1. ENTRA Pay Transparency Audit Q2 2026: ENTRA audited 87 AI companies against five dimensions of pay transparency (band disclosure, band width, equity compensation transparency, geographic consistency, retroactive disclosure). Twenty companies scored for the August 1 index release. Approximately 120 live postings reviewed per company where available; July 2026 audit window.
2. ENTRA Salary Survey Q2 2026 (n=2,140): AI practitioners across 47 countries. Compensation reported as total annual package including base, equity (annualized at grant price), and cash bonus. Survey fielded April–June 2026. Margin of error ±2.1 percentage points at 95% confidence.
3. ENTRA Job Signal Index Q2 2026: 47,200 active AI job postings analyzed for band disclosure rate, band width, and cross-jurisdiction consistency. Postings classified across 31 countries, April 1–June 30, 2026.
4. Levels.fyi Q1–Q2 2026 (n=3,800+ verified offers): External offer corpus for validation of ENTRA survey findings. Used for senior IC total compensation benchmarks at named employers.
All USD conversions use canonical rates: EUR/USD 1.09, GBP/USD 1.27, CHF/USD 1.085, SAR/USD 3.75 (pegged), BHD/USD 2.653.
Closing: the 2027 watchlist
Three leading indicators will determine whether the regulatory arc described in this report produces structural market change or a second generation of compliance theater.
Watch 1: EU enforcement Q1 2027. Slovakia, Estonia, Lithuania, and Malta, the four member states that transposed the EU Pay Transparency Directive on time, are where the first individual employer enforcement actions will occur. The first penalty assessed against a named AI employer under the Directive sets the market's estimate of enforcement risk for the 23 member states where transposition is still pending. An enforcement action against a mid-size AI employer in Bratislava or Tallinn will produce more behavioral change across the EU's non-transposing majority than another Commission infringement notice. Watch for enforcement actions from national labor inspectorates in those four jurisdictions in Q1 2027.
Watch 2: US federal PALS Act, Senate vote. The Pay Equity for Access to Lucrative Secrets Act (proposed, pending Senate vote as of Q3 2026) is the most consequential pending US legislation for AI compensation transparency. Senate Commerce Committee markup is scheduled for Q4 2026. If the bill advances to a floor vote and passes in its current form, the AI industry faces a federal disclosure obligation for the first time, closing the gap that state mandates leave open for employers in non-mandate states and below California's threshold. The Senate vote timeline will determine whether the federal framework leads or follows the 2027 regulatory cycle.
Watch 3: Saudi MOHRE Resolution 0340, Q1 2027 review. The first systematic compliance review of Resolution 0340 falls in Q1 2027. Saudi Arabia's Ministry of Human Resources and Social Development will assess whether international AI employers operating through Saudi entities met band posting requirements for Saudization-compliant roles. The outcome will determine whether Resolution 0340 operates as a real disclosure framework or a documentation formality. Given that Vision 2030 AI hiring targets require international employers to compete for Saudi national talent on terms that make compensation visible, the incentive to treat this review seriously is higher than the incentive to treat comparable EU obligations seriously in non-enforcement jurisdictions.
Forty-eight percent of AI job postings carried salary disclosure in Q2 2026. The number that matters is what that figure reads in February 2027: after the EU's first enforcement cycle, after the PALS Act markup, after the UK action plan deadline comes into view. ENTRA forecasts 74 percent of Fortune 500 AI postings globally will carry some form of disclosure by January 2027. The gap between 48 and 74 is where 87 employer decisions, 23 regulatory bodies, and 2,140 survey respondents' behavior aggregates into a market outcome. Whichever direction that gap closes from (candidates pulling toward transparency, regulators pushing from behind, or employers getting there ahead of both), the information architecture of AI hiring in 2027 will not look like 2026.
[^1]: ENTRA Job Signal Index Q2 2026. 47,200 active AI postings classified across 31 countries, April 1–June 30, 2026. Two disclosure figures are used in this report and are not contradictory: (a) 48% — postings carrying any form of salary disclosure, including indicative ranges, approximate bands, and jurisdiction-mandated numeric ranges (broader definition, used for hook and overall market characterization); (b) 41% — postings carrying a strict numeric floor-and-ceiling band only (narrower definition, used in comparative analysis and cited in ENTRA reports of August 7 and August 21). Both figures are editorially approved and sourced from the same ENTRA Job Signal Index Q2 2026 corpus. The distinction is defined in this footnote. ENTRA Q2 2026 Salary Survey: n=2,140 AI practitioners, 47 countries, fielded April–June 2026.
[^2]: ENTRA Pay Transparency Audit Q2 2026. 87-company cohort audited against five dimensions using ENTRA Pay Transparency Score framework. Camp assignments are ENTRA editorial classification based on Pay Transparency Audit Q2 2026 cohort (n=87). Camp thresholds: Discloser ≥70/100, Complier 40–69/100, Resister below 40/100. Percentage splits (24%, 47%, 29%) represent ENTRA's audit population, not the broader market. The August 1 ranking published scores for the top 20 companies within this cohort.
[^3]: Application volume premium (34%), screen-to-interview rates (68% vs. 49%), and offer acceptance rates (81% vs. 64%) from ENTRA Pay Transparency Audit Q2 2026 employer-reported data, cross-referenced against ENTRA Q2 2026 Salary Survey. Sunk cost per declined offer ($68,000) from ENTRA H1 2026 cost-of-acquisition model (frontier-lab senior IC tier). Annual cost figure: 17pp acceptance gap × 50 senior offers × $68,000 sunk cost = $578,000.
[^4]: Time-to-fill figures (Discloser: 46 days; Complier: 55 days; Resister: 68 days) from ENTRA Q2 2026 CHRO Survey supplementary data (n=44 CHRO respondents). BLS JOLTS June 2026: computer and information technology occupations, seasonally adjusted job openings (847,000).
[^5]: Anthropic band data (L5: $120,000–$320,000) from Q2 2026 California SB 1162-compliant posting audit. Band reflects posted floor and ceiling in active California-jurisdiction postings as of June 30, 2026.
[^6]: Anthropic total compensation ranges — L6 research: $480,000–$740,000; L6 product: $360,000–$540,000 — from ENTRA Q2 2026 Salary Survey (n=38 Anthropic respondents, seniority-filtered), corroborated against Levels.fyi Q2 2026 (n=3,800+ verified offers). Total comp includes base, annualized equity at grant price, and cash bonus.
[^7]: OpenAI score (65/100) and PPU-to-RSU conversion characterization from ENTRA Pay Transparency Audit Q2 2026 Equity Compensation Transparency dimension analysis. Conversion window: Q4 2025–Q1 2026, per ENTRA market tracking.
[^8]: Title fragmentation finding from ENTRA Job Signal Index Q2 2026 qualitative audit supplement. ENTRA reviewed posting title variations against O*NET occupational taxonomy to identify category-avoidance behavior in the Resister cohort.
[^9]: xAI audit score (28/100) and California posting characterization from ENTRA Pay Transparency Audit Q2 2026. California nexus determination based on ENTRA review of active Q2 2026 postings marked "US Remote" against the SB 1162 15-employee threshold applicability test. Data-provost note: xAI's compliance posture is ENTRA's editorial characterization based on publicly observable posting behavior. xAI has not been contacted for comment.
[^10]: Candidate uplift figures ($43,000 and $18,000) and offer-to-close rates from ENTRA Q2 2026 Salary Survey (n=2,140). Band percentile figures from survey sub-cohort with complete offer-acceptance records (n=1,180). Wide-band threshold: spread above $150,000. Tight-band threshold: spread below $80,000.
[^11]: Job abandonment rate (63%) from ENTRA Q2 2026 Salary Survey sub-cohort who encountered a band below their stated compensation floor (n=891). Question: "When the posted salary range was below your stated compensation floor, did you submit an application?" Response distribution: No, 63%; Yes, 3%; Contacted recruiter first, 34%.
[^12]: Constitutional AI disclosure rate (19%) and pay premium (34%); RLHF disclosure rate (31%) and pay premium (28%); multimodal systems and LLM inference optimization figures from ENTRA Q2 2026 Job Signal Index skill premium analysis, published in ENTRA Top 20 AI Skills by Pay Premium, August 25, 2026.
[^13]: Python/PyTorch baseline disclosure rate (71%) and commodity skill designation from ENTRA Q2 2026 Job Signal Index skill taxonomy.
[^14]: Geographic arbitrage finding (37%) from ENTRA Q2 2026 Salary Survey, US respondent sub-cohort in non-mandate states (n=412). Question: "Have you filtered job searches by state specifically to access posted salary bands?"
[^15]: EU transposition status (23 of 27 member states missed June 7, 2026 deadline; transposing states: Slovakia, Estonia, Lithuania, Malta) per ENTRA regulatory tracking as of August 28, 2026. 23 is the ENTRA-canonical figure as of August 2026; confirmed by Data Provost review (prior ENTRA article dated August 21 cited 12 — that figure has been superseded; August 10 briefing and August 24 weekly briefing both used 23). EU Directive 2023/970/EU Article 23 requires member states to establish effective, proportionate, and dissuasive penalties; the Directive does not specify a penalty percentage — enforcement frameworks and specific penalty thresholds are determined at the member-state level. No specific percentage is cited in this report. European Commission Article 258 infringement proceedings: EC press release, July 2026.
[^16]: Article 258 proceedings: European Commission press release, July 2026. CJEU referral timeline reflects standard infringement procedure sequencing; not a confirmed schedule for any specific proceeding.
[^17]: US federal PALS Act characterization based on ENTRA legislative tracking as of August 28, 2026. Act is proposed, not enacted; Senate Commerce Committee markup schedule is pending and not confirmed. CHRO sentiment figure (64%) from ENTRA Q2 2026 CHRO Survey (n=44 respondents).
[^18]: Illinois Equal Pay Act extension (threshold change to 100+ employees, effective January 1, 2027) per ENTRA legislative tracking. Chicago AI employer cluster characterization from ENTRA Q2 2026 Job Signal Index geographic classification.
[^19]: New Jersey SB1568 effective date (June 2026) and five-employee threshold per New Jersey Department of Labor regulatory guidance, per ENTRA tracking.
[^20]: UK Employment Rights Act 2025: Royal Assent December 18, 2025. Action plan threshold (250+ employees) and April 2027 deadline per Part 6 of the Act. ACAS/EHRC guidance status: draft consultation published June 2026; final guidance not issued as of August 28, 2026. Equity-as-percentage-of-total-comp figure (40–60%) for Google DeepMind London senior research roles from ENTRA Q2 2026 tracking (GBP/USD 1.27 canonical rate).
[^21]: Saudi MOHRE Resolution 0340 characterization based on ENTRA MENA regional reporting. 12-month implementation window and Q1 2027 review point per ENTRA regulatory tracking. Employers named per ENTRA Q2 2026 Job Signal Index MENA geographic cohort.
[^22]: ENTRA forecast (74% of Fortune 500 AI postings globally by January 2027) assumes EU full transposition by Q1 2027, continued US state mandate enforcement, UK ERA 2025 action plan compliance beginning April 2027, and no enacted US federal PALS Act. Estimate carries material uncertainty; actual coverage depends on enforcement pace and employer behavioral response.
[^23]: Anthropic L6 Research Scientist total compensation range ($480,000–$740,000) from ENTRA Q2 2026 Salary Survey (n=38 Anthropic respondents) and Levels.fyi Q2 2026. The 300-to-400 globally qualified Constitutional AI safety researcher figure is an ENTRA editorial estimate based on publication authorship analysis of frontier-lab safety research papers; it is not a surveyed count.
[^24]: CHRO Survey finding (71% reported increased internal compensation review requests within 90 days) from ENTRA Q2 2026 CHRO Survey (n=44). Question: "After your organization began posting salary bands, did you observe an increase in employee-initiated compensation review requests within 90 days of the first posting cycle?"
[^25]: California CRD 2025 Pay Data Report: mean unadjusted gender pay gap (17.4%) in professional computing occupations, technology-sector employers. Published May 2026, California Civil Rights Department. Within-category variance characterization is ENTRA editorial analysis applied to CRD published data.
ENTRA Job Signal Index, ENTRA Q2 2026 Salary Survey, and ENTRA CHRO Survey data are collected and analyzed independently by ENTRA Intelligence. No compensation is accepted from companies referenced in this analysis. Company-level compensation data are editorial estimates based on publicly available disclosure data, Levels.fyi self-report aggregates, and anonymized ENTRA survey respondents; they are not confirmed by the named companies and should not be treated as authoritative representations of those companies' compensation programs. Legal framework characterizations are editorial and do not constitute legal advice; employers should seek jurisdiction-specific counsel on disclosure obligations. Exchange rates: EUR/USD 1.09, GBP/USD 1.27, CHF/USD 1.085, SAR/USD 3.75 (pegged), BHD/USD 2.653.
Global Career PlatformFind AI talent. Find your next role.
Booking is hotels. · Airbnb is apartments. · ENTRA is global careers.