Industry Trends¶

Phase: 3 — Synthesis Project: likeness Date: 2026-05-09 Confidence: High on regulatory trajectory; Medium-High on technology and behavioral trends

Macro Trends¶

Direction of travel — federal and state level — is clearly toward mandated consent + provenance + identity verification for AI-generated media. Timeline: ENACTED through 2027.

Regulation	Status	Effective	Direct relevance to Likeness
TAKE IT DOWN Act (US federal)	Enacted May 2025	Now	48-hour platform takedown requirement (May 2026 compliance deadline)
California AB 2602 + AB 1836	Enacted Sept 2024	Now	Voids non-specific digital replica contracts; aligns with Likeness's license object
Tennessee ELVIS Act	Enacted July 2024	Now	First state AI likeness law; Class A misdemeanor framework
NO FAKES Act (US federal)	Pending	TBD	Federal right of publicity, lifetime + 70 years; "licensable not assignable"
EU AI Act Article 50	Active	Aug 2026 enforcement	Machine-readable disclosure on AI content (C2PA satisfies)
UK Online Safety Act (age verify)	Enforced	Jul 2025	Ofcom-grade fan-side age verification for UK launch
Mastercard adult content (AI extension)	Active	2024-2025 revisions	Written consent + age + identity for AI-generated content
Visa VIRP	Active	Apr 2024	<1% chargeback threshold; merchant policy review

Strategic implication: Likeness's architecture commitments (no model export, license-gated inference, per-creator isolation, per-output watermarking + signed metadata) are not just product features — they are pre-compliance with the direction of regulation. The platforms operating in opposition to this direction (Civitai pre-May-2025, gray-market deepfake sites) face escalating compliance friction.

Trend 2: Deepfake incident volume is exploding¶

500K → 8M+ deepfake incidents (2023 → 2025), +900% in two years [Tier 2]
~90% of deepfake content targets women [MIT Tech Review, Tier 1]
Public sentiment strongly negative on adult deepfake abuse (47.5% negative — highest of any deepfake topic) [Tier 2]
Adult AI-IBSA (image-based sexual abuse) is in the public consciousness as a category-defining harm

Strategic implication: Direct demand signal for a consent-first alternative. Cultural license for "we're the legitimate platform" framing is strong and growing.

Trend 3: Creator-side AI is being monetized¶

15% of Fanvue revenue is AI-generated content [Sacra, Tier 2] — fans demonstrably pay for AI on creator platforms.
AI companion apps overall: $120M projected 2025 revenue [Tier 2].
Adult AI image generation: $2.8B in 2025 (+340% from 2024) [Tier 3, directional].
337+ active revenue-generating AI companion apps as of July 2025; 128 new in last 6 months.

Strategic implication: Fans are not the question. Creator participation is the question. Demand for AI on creator platforms is established.

Loti AI: $22.85M total ($16.2M Series A April 2025, Khosla Ventures).
Vermillio AI: $16M Series A March 2025, Sony-backed.
CastNym: private beta.
Adult-creator-side equivalent: Vylit at $2.7M Seed, no others yet funded.

Strategic implication: The funding gap between mainstream consent infrastructure and adult-side consent infrastructure IS the opportunity. Likeness's pre-seed at $1.5M is appropriately sized.

Technology Shifts¶

Shift 1: Per-creator fine-tuning (LoRA / DreamBooth) is the primitive¶

Mature for hobbyist/gray-market use; legitimate-use adoption is early but credible.
Civitai received $5M from a16z in Nov 2023 specifically on this primitive.
Adoption stage: Growing rapidly.
Likeness implication: ML brief's architectural choice is correct.

Shift 2: Content provenance is becoming infrastructure¶

C2PA v2.2 published May 2025; v2.3 in draft.
Samsung Galaxy S25 ships with C2PA on AI-edited photos (Jan 2025).
CISA (Jan 2025) advisory recommends C2PA for government / critical infrastructure.
EU AI Act Article 50 enforcement Aug 2026.
Adoption stage: Standards-led, mainstream within 12-24 months.
Likeness implication: Adopt C2PA from day 1; consider C2PA membership.

Shift 3: Watermarking is moving from research to standard¶

Tree-Ring, StableSignature, and successor invisible watermarking schemes are production-ready for image content.
Combined with C2PA manifests = "Durable Content Credentials" (the C2PA-recommended pattern).
Adoption stage: Production-ready for image; less mature for video.
Likeness implication: ML brief's "watermark + perceptual hash + signed metadata" stack is the right shape. Build on the standards rather than inventing.

Shift 4: Distilled / accelerated model variants reduce inference cost¶

Flux Schnell, SDXL Turbo, Hyper-SDXL etc. enable real-time-ish generation at 5-10x lower compute cost vs. full-quality.
Likeness implication: Cost / latency / quality trade-off curve is real and non-trivial. ML brief notes this; ML Lead must own the operating point.

Investment & M&A Signals¶

Round	Date	Significance
Loti AI $16.2M Series A	Apr 2025	Mainstream consent infrastructure has tier-1 VC commitment
Vermillio AI $16M Series A	Mar 2025	Music industry vertical of consent infrastructure
Fanvue $22M Series A	Jan 2026	Creator subscription platform with AI integration is investable
Vylit $2.7M Seed	Sep 2025	Adult-creator-side consent infrastructure is investable; OnlyFans-pedigree founder commands attention
Civitai $5M	Nov 2023	a16z-backed LoRA marketplace; subsequently de-banked over content concerns

Smart money trajectory: Mainstream consent infra has cleared Series A scale. Adult-side equivalents are at Seed scale, with Vylit being the bellwether. A successful Likeness Series A (post-MVP) is plausibly $10-20M in 18-24 months based on Loti/Vermillio comparables.

Behavioral Shifts¶

Creator side¶

Active labor movement against perpetual likeness contracts in adult industry venues. California AB 2602's framing was specifically designed to constrain studio-extracted likeness transfers.
Mid-tier creators (top 1-10% of OnlyFans) face revenue ceilings on real content; AI is a potential additive line.
Creator concern about AI eroding fan trust is named in industry coverage.
Adoption of AI tools by creators is starting (chatbots for fan response, AI-assisted content) but is uneven.

Fan side¶

Willingness to pay for AI-generated adult content is established ($120M in AI companion apps 2025).
Existing creator-platform fans demonstrably pay for AI of their existing creator (Fanvue 15%).
Fan migration friction across platforms is real but soft; fans follow creators.

Regulatory Trajectory¶

Direction of travel: tighter, more consent-required, more identity-required, more provenance-required.

The next 12-24 months will likely include: - NO FAKES Act enactment (or near-equivalent state-level expansion) - More state-level AI likeness laws (every state introduced legislation in 2025) - Continued tightening of Mastercard / Visa rules on adult-content category - EU AI Act Article 50 enforcement begins - UK Ofcom expanded enforcement waves

Likeness implication: Build for the regulatory floor of 2027, not the floor of 2026. The compliance investment pays compounding dividends as competitors are forced to retrofit.

What This Means for Likeness — Strategic Implications¶

Architectural commitments are positioning. No model export, license-gated inference, per-creator isolation are not "extras" — they are the consent infrastructure that the regulatory and processor environments increasingly require. Tell this story aggressively.
The window is real but finite. Vylit launched in 2026. Fanvue could pivot. A team with adult-industry insider trust + ML capability could ship a competing platform in 12-18 months. Speed of cofounder hires (Creator Ops, CEO) is competitively load-bearing.
Consent-first explicit-allowed is the uncontested cell. Vylit chose explicit-OUT. OnlyFans + Fanvue have the platforms but lack architectural commitments. Likeness's combination is unmatched today.
Demand is established; creator participation is the bet. Fans pay for AI on creator platforms. Whether creators participate in the specific configuration Likeness offers is the load-bearing unknown — and it's a discovery question, not a market-trends question.
C2PA membership is a low-cost, high-credibility positioning move. Pre-launch, ideally.

Timing Scorecard¶

Tailwind	Strength
TAKE IT DOWN Act enacted	High
California AB 2602 in effect	High
C2PA standardization	High
EU AI Act Article 50 enforcement	Medium-High
Fans pay for AI on creator platforms	High
Mainstream consent infra well-funded; adult-side gap	High

Headwind	Strength
Payment processor risk (Civitai precedent)	Medium-High
Vylit launched and well-funded	Medium
Fanvue could pivot	Medium
337+ adult-AI-generator fragmentation	Low-Medium

Net assessment: Strong tailwinds, real headwinds. Now is a good time to build, with execution risk concentrated in cofounder hires and creator validation.

Data Gaps¶

Specific NO FAKES Act passage timeline is uncertain — could pass 2026 or 2027 or stall. Position the bet to win in either case.
Fanvue's specific roadmap re: real-creator AI licensing is unknown.
Mastercard / Visa AI-specific rule trajectory — public statements have been broad; specific tightening is unpredictable.

Strategic Connections¶

The regulatory tailwinds in this document support the architectural-commitments differentiation in competitor-landscape.md.
The "window is finite" signal connects to 06-validation/validation-playbook.md — speed of validation is competitively constrained.
The behavioral shift evidence (Fanvue 15% AI revenue) supports the business model assumptions in 02-strategy/lean-canvas.md and 05-financial/revenue-model.md.

Flags¶

Red Flags: - None at the trends level.

Yellow Flags: - Regulatory trajectory is favorable but not deterministic. NO FAKES Act could stall. Position language on regulatory tailwinds carefully — describe direction, not specific outcomes. - The 900% growth in deepfake incidents is a category-level harm signal that benefits Likeness's positioning AND raises platform-level compliance bar for everyone, including Likeness.

Sources¶

See 01-discovery/raw/trends.md, raw/regulatory.md, raw/adjacent-markets.md for full citations.