Written from 16 named sources · Nov 1 · first result Forensic Due Diligence Report: Hax Labs (hax-labs.com) Target Entity: Hax Labs (hax-labs.com / The Haxakain Group) [15] Operating Segments: Hax Voices (Talent-Facing Marketplace & Agency) Hax Vault (Enterprise Voice Deployment & AI Agents) [2, 8] Evaluation Practice: Voice-Over Talent Representation, Biometric Rights Governance, Talent Compensation & Contract Benchmarking Assessment Date: November 1, 2026 — Evidence reflects supplied page snapshots gathered earlier; live-site parity as of today cannot be confirmed. Executive Summary & Due Diligence Dashboard Hax Labs markets itself as an ethical voice-talent agency and production entity tailored for the synthetic-speech era [2, 8]. Across its public touchpoints, the platform promotes "licensed human artists," human-first AI infrastructure, and strict talent consent [2, 8]. The platform is segmented into two operational divisions: Hax Voices, aimed at voice-over talent seeking automated marketplace monetization without studio requirements [2], and Hax Vault, an enterprise deployment division delivering custom AI voice agents, corporate e-learning narration, and business call-center integration [2, 8]. Despite promotional messaging referencing ethical AI and the National Association of Voice Actors (NAVA) [7], review of the supplied public materials reveals material contractual ambiguities, non-transparent talent compensation, and structural intellectual property (IP) risks [2, 8]. Foremost is tension between talent likeness rights and enterprise marketing: Hax Labs promises business buyers a proprietary model that can be “owned by your company” under the slogan “One Voice. Your Brand. Forever.” [8], while simultaneously listing “ongoing licensing” for the same product — leaving whether the buyer receives ownership, an exclusive license, or a managed service unresolved on the public record [8]. Table 1: Diligence Confidence & Risk Index Assessment Area Evidence Level Rating / Verdict Primary Risk Indicator Talent Affiliation & Provenance Low to Moderate High Risk Pseudonymous listings; no legal names, union, agency or credits disclosed on supplied pages [2, 14]. “14+” claimed on [8] against 13 names on [2]. Compensation Transparency Opaque on supplied pages Critical Risk No talent payout split, capture stipend, minimum guarantee or audit term on [2], [8], [14]. “Revenue details are temporarily unavailable” on /talent [14]. Apply-flow contract not supplied. Contractual Alignment Contradictory marketing; incomplete record High Risk “Owned by your company” + “Forever” sits alongside “ongoing licensing” and undefined “licensed” [8]. NAVA mention verified [7]; adoption of NAVA AI Rider unsubstantiated. Enterprise Pricing Transparency Defined starting points Moderate Risk Published bands mask talent pass-through: $2,500–$8,000 + $800–$3,000/mo; $5,000–$25,000 + ongoing licensing; $150–$350/finished hour [8]. Exact pricing confirmed only on discovery call [8]. Acoustic & Production Quality Self-Reported / Selective Moderate Risk Roster audition blocked by “temporarily unavailable” cards [2, 14]; no MOS, latency or WER benchmarks published [8]. “Zero downtime” and “2B+” unevidenced [8]. Scope limit: Sources [1], [3] and [6] support nothing here. Terms, MSA, full privacy policy, and apply-flow contract were not available, so absence findings apply only to pages supplied [2, 7, 8, 14]. Pillar 1: Voice Talent Affiliation & Roster Verification Audit Hax Labs publicly presents 13 distinct names as of capture: Bella E., Cherie R., Jules R., François-Louis, Lennox, Ingrid F., Miguel G., Kleo, Mark D., Elijah S., Brian, Jedediah, and Kathy [2]. The [14] capture omits François-Louis in its first carousel row only, but otherwise mirrors the set [14]. Key Analytical Findings: Talent Pseudonymity and Verification Deficits: Only first names or initials appear, with no legal names, union affiliation, agency representation or verifiable credits. The page asserts “Real voices. Real artists.” without proof [2, 14]. This prevents outside verification of whether performers are career voice-over artists, freelance hires, or one-off buyouts [2, 14]. The “14+ licensed human voices” claim on [8] is inconsistent with 13 names listed on [2]. While reels, credits and union membership (SAG-AFTRA, ACTRA, UK Equity) are common vetting tools, verification is not uniform and much talent is non-union — “routinely verified” would overstate practice. Founder Model (“Brian”) — Self-Reported Only: Founder and CEO Hank Groff is identified as such on [2] and [13], with company start listed as Dec 2024 on [13]. The voice-career narrative rests entirely on self-published material: Hax states a “leading AI voice platform… engineered it into… the default voice ‘Brian’” [2], and Groff’s LinkedIn states “top voice on ElevenLabs (the iconic ‘Brian’ voice)” [13]. No independent confirmation of the voice career, “Brian” provenance, or the six Brian demo videos (Insurify/Disney+, Elon Musk interview, etc.) is in the evidence. Hax pages do not name ElevenLabs — only Groff’s LinkedIn does [13] — and describe only a “partnership” with no terms [2]. Sales / Coaching History — Self-Reported Only: [13] lists dealer-division and CRO roles (Guardian, Dynamark) and a “Business Coach” headline; [5] states “25 years of dealer network experience.” All are self-published and not independently verified [5, 13]. Aggregator Distribution Model (Inference): Hax Voices states: “We build your AI voice character, place it on leading marketplaces, and split the revenue fairly. No studio required.” [2]. This supports third-party marketplace placement with an undefined “fairly” split, plus the earnings disclaimer “Results are not typical” / “Individual results vary” [2]. Whether Hax functions as exclusive agency vs. non-exclusive aggregator / licensing clearinghouse is an inference — exclusivity terms are not documented. The “#1 ethical voice talent agency” [2] and “200+” newsletter [2] claims are unevidenced superlatives. Contact is info@haxakain.com with Hax Labs under The Haxakain Group, leaving the contracting entity / licensor of record unclear [8, 15]. Pillar 2: Compensation Models & Rate Card Benchmarking In professional voice-over and ethical synthetic licensing, compensation is typically staged across training-session fees, generative-usage royalties, term-limited windows, and category exclusivity — but no supplied Hax page defines these for talent. Enterprise Pricing Disclosures vs. Talent Compensation Hax Labs publishes “Transparent Starting Points” without requiring sales inquiry, while stating exact pricing is confirmed on discovery call and enterprise volume deals are scoped by Groff [8]: AI Voice Agent Deployment: $2,500 – $8,000 setup + $800 – $3,000 / mo management [8]. The retainer is described as “flat monthly management fee only” after setup for voice agents — not for custom builds or narration [8]. Custom Enterprise Voice Build: $5,000 – $25,000 build + ongoing licensing [8]. Corporate Narration: $150 – $350 / finished hour, volume discounts available [8], plus $1,500 pilot module, from $4,800/yr for 4 modules, and from $14,400/yr for 12+ modules [8]. Module length is undefined, so per-hour vs. per-module reconciliation is not possible. Multilingual Deployment: Custom quote [8]. No split, stipend or minimum guarantee appears on [2], [8], [14] or [16]. On /talent, “Revenue details are temporarily unavailable” suggests a revenue panel may exist but does not render [14]. The apply-flow contract was not seen, so talent economics cannot be confirmed. Table 3: Economic Benchmarking (Hax Labs vs. Industry Baselines — With Evidence Caveats) Parameter Common Industry Talking Points Hax Labs Observed Model Compliance / Diligence Assessment Initial Capture / Model Training Fee $1,500–$7,500+ for 2–4 hours of source audio is often cited, but no supplied source verifies this and AI capture fees vary widely, including under union rules. ~30 min to a few hours is plausible but unproven. Undisclosed on supplied pages. Enterprise billed $2,500–$8,000 setup; $5,000–$25,000 build [8]. Critical gap (pages only): No confirmation a dedicated training fee is passed through [8]. Residual / Compute Royalty Rate Per-minute / per-word renewals with floors are best-practice concepts; a 20–50% marketplace share is plausible but unevidenced in supplied sources. The [4] snippet (~$250/mo, day of self-recording) does not contain a $0.03–$0.05 per 1,000 characters rate, and “Reader Library” terminology is questionable [4]. Undisclosed. “Split the revenue fairly” — “fairly” undefined [2]. High Risk: No defined schedule, floor, or audit right on supplied pages [2, 8]. Enterprise Management Allocation Licensors often seek a share of recurring licensing, but “demand 20–50%” is loosely worded and unevidenced. Enterprise billed: $800–$3,000/mo (agents) [8]. Talent split: undisclosed [8]. Custom builds list “ongoing licensing” separately with no talent share stated [8]. Arbitrage risk (conditional, undocumented): If “management” is treated as infrastructure, not likeness license, talent may see no residual — but this is speculative without the contract [8]. E-Learning Finished Audio Rates No supplied source gives a $750–$1,500+/hr GVAA corporate benchmark; GVAA guidance is typically per finished minute/word, making per-hour comparison unreliable. Hyperscaler neural TTS is roughly an order of magnitude cheaper than $0.15–$0.30/min premium API pricing — the “major providers” generalisation is false. Enterprise billed: $150–$350 per finished hour [8]. Talent split: undisclosed [8]. Deficit flag: Billed rate sits well below typical human non-broadcast narration expectations, signaling commoditized synthetic pricing — merits scrutiny, not proof of underpayment [8]. Forensic Compensation Risk: Margin Decoupling (Conditional) For agents, [8] calls the recurring fee “management” only; for custom builds it lists “ongoing licensing” separately but never states whether talent shares in it [8]. Talent should therefore require contractually defined dollar rates for onboarding, fixed percentages of setup/build, and an auditable percentage of recurring fees — and refuse vague “fair” language [2, 8]. Pillar 3: Contractual Alignment & Biometric Rights Governance The Corporate “Ownership” Contradiction — Verified Language, Ambiguous Legal Effect Hax Labs states: “One Voice. Your Brand. Forever.” and “proprietary AI voice trained on a licensed human artist — owned by your company” [8]. “Forever” is presented as brand-voice slogan, not a stated license term, and sits in tension with “ongoing licensing” for the same build [8]. “Licensed” itself is never defined — no term, scope or exclusivity [8]. Legal exposure must be stated carefully: Perpetual or irrevocable transfer of voice/likeness is disfavoured and risky, but “cannot assign” is too absolute — rights of publicity are generally licensable, with newer statutes (e.g., California AB 2602, Tennessee ELVIS Act, NY digital-replica provisions) and SAG-AFTRA/NAVA guidance restricting broad, vague or perpetual terms and requiring informed consent. Voice as such is not copyrightable (cf. Midler, Waits theories); protection is mainly state law, plus Lanham Act false endorsement and BIPA in some contexts. There is no general federal right of publicity statute; federal bills cannot be confirmed as enacted on this record. Buyers using replicas beyond granted scope can face publicity and consumer-protection claims — appropriately hedged. Lehrman v. Lovo (SDNY 2025) allowed such state-law claims to proceed. A licensee cannot convey more than it holds — sound as conditional logic — but the factual predicate (that Hax talent terms are non-exclusive or revocable) is unseen. Enterprise buyers therefore face potential defective chain-of-title if Hax markets “ownership” while holding only a limited license [8]. NAVA AI Rider Alignment — Mention Verified, Adoption Unsubstantiated [7] contains the text “NAVA — National Association of Voice Actors” next to “Compliance,” but the snippet gives no context [7]. A mention is not adoption; “internal policy architecture” is gloss, not evidence. Source [9] is a talent’s personal PDF (“My voice is not to be used or sold for Ai training”) — it does not show NAVA rider text or an April 14, 2026 update date. A NAVA AI rider with separate-consent language exists in general training knowledge, but exact wording and date are unconfirmed and must not be quoted as verified. On the supplied record, [8] has no prohibition on secondary fine-tuning, no weight-destruction protocol, and no output-harvesting restriction. Because no Terms or MSA was supplied, these are absence findings for [8] only, not comprehensive contract conclusions. Competitor Platform Restrictive Covenants — Narrowed Correction Hax names exactly: “OpenAI · Speechify · ResembleAI · PlayHT · Respeecher · Murf · HeyGen · RaskAI” as “Direct Competitors” [8] and states “any deal involving them requires direct review by Hank Groff before signing” [8]. Crucially, [8] adds: “This restriction applies to Hank's personal voice only — other roster artists may serve these platforms subject to individual review.” There is residual ambiguity between deal-level review and voice-level restriction, but no blanket exclusivity on other artists is documented — the broader exclusivity concern in the prior draft is unsubstantiated and partly contradicted [8]. Pillar 4: Production & Acoustic Quality Assessment Hax claims “No synthetic TTS artifacts. No robotic delivery.” and “Brand Consistent” [8]. The phrase “natural conversational delivery” was not found in supplied text. These are marketing, not guarantees. The page contrasts its voices with “Synthetic neural net” TTS while selling AI voice agents — misleading framing [8]. Technical Findings & Quality Deficits (Corrected): Roster Outages — Precise Attribution: [2] shows “The voice roster is temporarily unavailable” and “Voice temporarily unavailable” on six descriptor cards (“Deep, Rich & Cinematic,” “Youthful Energetic Social,” “Warm, French-Accented Prestige,” “Smooth, Professional Delivery,” etc.) [2]. The descriptors are verified, but descriptor-to-persona mapping is unsubstantiated — banners are not attached to named personas (Bella E., Kathy, etc.). The “Revenue details are temporarily unavailable” banner appears on /talent [14], not on /creators [2]. Current status is untestable. Auditioning Impeded, Not Prevented: Roster samples are down, but six Brian demo videos are present [2], plus “Corporate Voice Samples” (Hannah, Hank Conversational) with 10+ more voices offered on [8]. Marketplace listings are another channel. Auditioning is impeded for the roster but not wholly prevented [2, 8]. Technology Stack — Undocumented Inference: No documentation of stack. “Credit Characters Generated” is ElevenLabs-style terminology, which is only suggestive [8]. References to Azure or Cartesia would be without basis. White-labelling at an “agency premium” is unevidenced; indeed $150–$350/finished hour looks inexpensive against human benchmarks [8]. At most, use of established commercial foundation-model APIs under the hood is a fair, flagged inference — not a finding. Latency vs. Fidelity — Outdated Framing Corrected: [8] markets “AI Voice Agents… call center agents. 24/7 availability, zero hold times,” deployable in about 2 weeks, with no technical features such as interruption handling or latency given [8]. A real latency-fidelity tension exists, but “multi-second diffusion vs. <300ms robotic IVR” is outdated: streaming low-latency TTS now delivers high quality well under 300ms time-to-first-audio, with conversational targets normally ~500–800ms end-to-end. No Independent Benchmarks: No MOS, latency or WER figures appear [8]. “Zero downtime” and “24/7” are unevidenced [8]. Weight portability is silent while claiming “owned by your company” plus “ongoing licensing” [8] — if the stack is third-party, weights are unlikely exportable. Lock-in is an inference from the documentation gap, not an evidenced restriction. Actionable Risk Matrix & Procurement Blueprint Table 4: Risk Matrix for Enterprise Procurement & Talent Partners Risk Category Inherent Risk Residual Risk Evaluative Finding / Diligence Recommendation Chain of Title & Likeness Rights Critical High “Owned” + “Forever” vs. “ongoing licensing” and undefined “licensed” creates title ambiguity [8]. Demand indemnification and inspection of redacted talent consent; clarify exclusive license vs. assignment under applicable state publicity / digital-replica law. Compensation Arbitrage High High No talent split, MG or audit term on supplied pages [2, 8]; revenue panel non-rendering on [14]. Contract unseen. Require dollar rates and accounting rights before recording. Platform Lock-In & Exclusivity Moderate Moderate No weight-portability term; “owned” language without export right [8]. Competitor review applies to Hank’s voice only; others subject to individual review — confirm in writing [8]. Require SLA: TTFB, uptime, pronunciation support. Technical Provenance Moderate Low Stack undisclosed; ElevenLabs named only on LinkedIn [13], not Hax pages. Verify deployment infrastructure, data quarantine (no retraining on call data/scripts), and whether samples are live vs. generated [8]. Entity & Key-Person Moderate Moderate Contracting entity unclear (Hax Labs vs. Haxakain Group, info@haxakain.com) [8, 15]; Groff personally reviews all enterprise engagements [8]. Confirm licensor of record and succession. Stakeholder Decision Checklists For Professional Voice-Over Talent (Evaluating Hax Voices Intake): [ ] Contract Entity Identification: Require exact legal entity (Hax Labs vs. The Haxakain Group or affiliate) and licensor of record [2, 13, 15]. [ ] NAVA AI Rider Incorporation: Do not rely on the [7] mention. Require the complete, current NAVA AI Rider as a schedule, expressly prohibiting foundation-model training and unapproved sublicensing — reviewed by counsel. [ ] Deconstruct “Ownership” Claims: Grant only an exclusive/non-exclusive time-delimited license; strike any term allowing marketing of your replica as “owned by” clients or “Forever” without renewal [8]. [ ] Transparent Royalty Framework: Refuse undefined “fair revenue split” [2]. Insist on onboarding fee, fixed % of setup/build, auditable % of monthly management / ongoing licensing, payment timing, and accounting [8]. [ ] Takedown & Destruction Verification: Mandate term, territory, use-case limits, de-indexing and verified weight destruction across distributed marketplaces on expiry/termination. Note: full Terms not reviewed — verify directly. For Enterprise Procurement Teams (Evaluating Hax Vault Engagements): [ ] Audit the Underlying Artist Agreement: Before $5,000–$25,000 builds, demand redacted talent license confirming informed consent for interactive agent / e-learning deployment [8]. [ ] Verify Title and IP Claims: Clarify “owned by your company” [8] — exclusive license to fine-tuned checkpoint vs. likeness purchase — and reconcile with “ongoing licensing” [8] under counsel review. [ ] Technical Performance SLAs: Require verifiable TTFB, >99.9% uptime (not unevidenced “zero downtime” [8]), pronunciation correction, revision and LMS-delivery obligations [8]. [ ] Data Quarantine Covenants: Secure binding warranties that live call data and proprietary scripts are never ingested for retraining [8]. Confirm pilot ($1,500) vs. annual blocks ($4,800 / $14,400) scope, module length, and volume-discount mechanics [8]. Sources [1] 2026 Crypto Crime Report – Illicit Crypto Trends & Typologies TRM Labs — https://www.trmlabs.com/reports-and-whitepapers/2026-crypto-crime-report [2] Hax Labs Advanced AI Voice Talent — https://hax-labs.com/creators [3] Comcast — https://en.wikipedia.org/wiki/Comcast [4] I'm earning $250/month fully passive income with ElevenLabs - Reddit — https://www.reddit.com/r/passive_income/comments/1ggaqhh/im_earning_250month_fully_passive_income_with [5] Hax Growth - Zero-Touch AI Sales Engine The Haxakain Group — https://www.hax-growth.com [6] pdf.pdf — https://www.statestreet.com/web/about/our-story/annual-report/documents/2025-annual-report.pdf [7] Privacy Policy - Hax Labs Advanced AI Voice Talent — https://hax-labs.com/privacy [8] For Business — AI Voice Agents & Enterprise Voice Hax Labs — https://hax-labs.com/for-business-overview [9] No Ai! — https://lanamarielive.com/audio/noai.pdf [13] Hank Groff - Hax Labs (hax-labs.com) — https://www.linkedin.com/in/hankgroff [14] Hax Labs Advanced AI Voice Talent — https://www.hax-labs.com/talent [15] The Haxakain Group Command Center — https://haxakain.com [16] Hax Labs Advanced AI Voice Talent — https://www.hax-labs.com/ AI Rider - NAVA — https://navavoices.org/ai-rider Hax Labs Advanced AI Voice Talent — https://hax-labs.com Hax Labs Advanced AI Voice Talent — https://hax-labs.com/library