Chapter 10: Global Cosmetic Firms and AI Microbiome — L'Oréal, Unilever, Estée Lauder, Shiseido, Amorepacific, LG H&H

Why this chapter

Brief A explicitly asked for "the concrete cases in which global firms have actually used AI to produce results." Across nine chapters we have followed the mechanism (Part I), the AI inflection (Part II), and the synthetic-biology / clinical bridge (Part III). Chapter 10 is the first chapter that lands all of that technology on specific companies. Academic flows look similar everywhere; industrial flows diverge under the pressures of capital, data, regulation, and brand strategy. The Big-4 (L'Oréal, Unilever, Estée Lauder, Shiseido) are similarly sized in global revenue but have wildly different shapes of AI-microbiome bets — and the Korean industry sends a unique signal in which the OEM (COSMAX) discloses more than the B2C brand owners (Amorepacific, LG H&H).

The aim is to give a decision-maker enough surface area to debate, in their own R&D meeting, why does our company not bet the way that company bets? So we repeat a four-axis frame for each firm — (1) data assets, (2) AI stack, (3) microbiome IP / products, (4) externally auditable disclosure. The last axis is the deepest message of this chapter — real industrial activity and disclosed activity are not the same, and interpreting non-disclosure (especially from Amorepacific and LG H&H) as non-activity will produce the wrong R&D decisions (Gap 4).

The three quantitative anchors of this chapter 1. The Big-4 data-asset gap: Unilever's self-reported skin-microbiome corpus of ~30,000 samples / ~5 billion data points across all major body sites ^[29] is larger than any academic cohort externally known. L'Oréal claims 20 years of in-house microbiome R&D plus the Lactobio acquisition's ~10,000 characterized isolates ^[21]. Both firms gate the data itself as core IP. 2. A Korean OEM operates the only peer-reviewed cohort comparable to Big-4: the COSMAX × Dankook FACE-LINK platform sits on a ~1,000-Korean cohort ^[24] and is one of the very few industrial cases with peer-reviewed output in Frontiers in Cellular and Infection Microbiology. Amorepacific and LG H&H each show effectively zero peer-reviewed outputs — and that should be read as a disclosure-policy difference, not an activity difference (Gap 4). 3. As of 2026-05, the count of peer-reviewed clinical outputs for an AI-designed cosmetic active is still zero. This is confirmed directly by L'Oréal-co-authored PRISMA systematic review ^[16], so Chapter 5 / Chapter 9's Gap 1 falls intact onto Chapter 10. The distance between "we used AI" as external marketing and "AI's contribution was peer-reviewed" remains large.

10.1 Methodology note — what we saw and what we did not

This chapter uses only publicly disclosed material — peer-reviewed papers, filed and published patents, official press releases, product launches, analyst-day decks, and publicly searchable IR documents on KIPRIS / USPTO / EPO. Internal R&D reports, undisclosed datasets, internal digital-twin architectures, and private clinical data are out of scope. This limit is decisive — the cosmetic industry deliberately holds large portions of R&D depth private, so the cross-company comparison here is fundamentally a comparison of disclosed depth, not of actual depth.

The asymmetry runs in both directions. Western majors are aggressive on stages like R&I editorials, VivaTech, SXSW, CES, and their content can make R&D depth look larger than it is. Korean majors (especially Amorepacific and LG H&H) center on KIPRIS patents, KSCS conference talks, and Korean-language trade press; to an English-only reader, their R&D depth can look smaller than it is. In both directions, the picture an outside observer sees is not the actual picture.

Operating rules of this chapter — for each company we (i) cite disclosed quantitative metrics as-is and flag the absence of external validation, (ii) prefer peer-reviewed output when it exists, (iii) explicitly label press-grade claims as such, (iv) treat large information asymmetry as a data point in its own right. We apply directly what ^[4] noted — "most cited efficacy claims rely on company press releases, not peer-reviewed clinical trials."

10.2 L'Oréal — the Big-4 player most aggressively integrating the AI-microbiome stack

In 2024-2025 L'Oréal is the Big-4 player that most aggressively markets AI + microbiome + device as one integrated stack. The skeleton in a sentence — research substrate (Lactobio + 20 years of in-house microbiome R&D) + AI diagnostic layer (Skin Genius, Cell BioPrint) + AR / device layer (Modiface, Modjoul Beauty Tech, AirPick) + global PR stages (VivaTech, CES). We take each in turn.

Research substrate. L'Oréal announced the acquisition of Copenhagen-based precision-probiotic and microbiome research firm Lactobio in December 2023 (closing into Q1 2024) and integrated ~10,000 characterized isolates with associated efficacy data into in-house R&I ^[20]. This is more than a library purchase — Lactobio's value proposition is strain × efficacy-labeled learning substrate, which is the rarest form of data an AI model needs when learning to predict efficacy candidates (cf. Chapter 4's data-asset discussion). The post-acquisition R&I editorial "The Future of Cosmetics Is Playing Out In The Microbiome" ^[22] explicitly maps a five-year microbiome R&D priority. L'Oréal's microbiome R&D also rests on older external partnerships — the 20-year collaboration with UC San Diego's Center for Microbiome Innovation has supplied an academic substrate ^[22].

AI diagnostic layer. Skin Genius is L'Oréal's best-known consumer-facing AI diagnostic, rolled out to 28 countries ^[21]; from a single selfie it scores 13 skin attributes and maps recommendations against an internal formulation database. Cell BioPrint is the home-diagnostic device unveiled at CES 2025 ^[23], estimating collagen biomarkers and recommending a regimen. Both tools are one half of a diagnostic-recommendation closed loop — the other half is the efficacy model that lives inside the internal formulation database, and that architecture is not disclosed.

AR and device layer. Modiface (acquired 2018) supplies the AR visualization layer; Modjoul Beauty Tech unveiled a family of AI devices at CES 2025 including Air Light Pro (low-temperature hair dryer, ~30% energy savings claim ^[23]) and AirPick (scalp-microbiome care device). The scalp microbiome device makes claimed efficacy assertions with no disclosed clinical validation — a clean instance of the broader Big-4 pattern in which device-side marketing outruns clinical output.

Auditable outputs — Haykal 2025 and Gueniche 2022. L'Oréal's most credible disclosed AI-microbiome outputs are two peer-reviewed reviews. (i) ^[15] is an R&D internal review whose ten authors are all L'Oréal-affiliated; it honestly enumerates the analytical toolbox the firm uses (16S, shotgun, metabolomics, ex vivo organotypic skin) and the limits it has hit internally (low biomass, batch effects, sample-collection variability). (ii) ^[16] is an L'Oréal R&I co-authored PRISMA-2020 systematic review that screened 403 papers, included 74, and explicitly stated — as of November 2025 there is no peer-reviewed clinical output for an AI-designed cosmetic active. That L'Oréal co-authored this conclusion is the most honest data point in this chapter — fast internally, thin in externally auditable output.

The trap of the "Big-4 leader" framing. L'Oréal is often called the Big-4 microbiome-AI leader because of integrated-stack visibility (Lactobio + Skin Genius + Cell BioPrint + Modiface). But that stack's depth is not disclosed with the same methodological backbone as Unilever × IBM's Carrieri 2021 peer-reviewed case (Chapter 8). It is more honest to read the asymmetry as L'Oréal exhibits more; Unilever documents more.

10.3 Unilever — the only Big-4 to disclose operational KPIs

Unilever is the only Big-4 firm to have publicly disclosed operational KPIs in quantitative form. In one paragraph — Unilever Beauty & Wellbeing R&D self-reports the largest human skin-microbiome dataset (~30,000 samples / ~5 billion data points, all major body sites), more than 100 microbiome patents filed, AI virtual cohorts of 2,500 simulated subjects, and the cycle KPIs consumer insight 60% faster / formulation cycles 5–6 → 1–2 / claims generation 75% quicker ^[29].

Data asset. The 30K-sample corpus is larger than any academic cohort we saw in Chapter 3 (an order of magnitude above HMP skin records). The "5 billion data points" figure is computed on sample × OTU × metadata dimensions, and this corpus is — if released — exactly the substrate size that could train the skin-microbiome foundation model named in Gap 2. It has not been released. That asymmetry is one of the core arguments behind the dataset-cooperative proposal in Chapter 12.

Carrieri 2021 — the methodological backbone of the industry. Unilever's most credible output is ^[6] — the IBM Research × Unilever R&D collaboration published in Scientific Reports. The methodology we detailed in Chapter 8 (gradient boosting + SHAP, Canadian + UK validation cohorts, hydration AUC ~0.7–0.8, menopause AUC ~0.85, smoking AUC ~0.75) has become the de facto standard the cosmetic industry uses when it wants AI to be interpretable. SHAP attribution is the core because — as Chapter 8 argued — cosmetic claim substantiation in front of legal review requires interpretable feature attribution.

POND'S 60-minute in-store diagnostic. ^[26] announced the first mass-market in-store skin-microbiome analysis → personalized recommendation device, with a ~60-minute turnaround. It is the most honest consumer-grade microbiome diagnostic shipped by any Big-4. External validation is not disclosed, and the specification (point-of-care PCR or fast-turnaround amplicon-only — likely candidates) is not public.

Honest reading of the operational KPIs. 5–6 → 1–2 formulation cycles and 75% faster claims generation are self-reported KPIs without external audit (Gap 15). They are nonetheless more credible than typical press-release claims for two reasons — (i) Unilever repeats the KPI consistently across SXSW 2025 ^[29] and a Scientific Reports paper ^[6]; that consistency is rarer in fabricated numbers; (ii) Unilever alone has the data asset (30K samples) that could plausibly support such a cycle compression, and an outside observer can partially verify that asset's existence. The honest read is the KPIs are plausible but unaudited, so discount them to roughly 1.5–2× cycle compression rather than the ~3× the press release implies.

Skin-brain axis expansion. Unilever publicly disclosed skin microbiome ↔ mental wellbeing research in 2025 ^[32], an experimental bet that extends the gut-brain-axis paradigm (Chapter 11) into skin. The 2026 forward outlook ^[31] codifies this direction into a five-year roadmap. Whether it maps onto clinical efficacy is an Open Question.

10.4 Estée Lauder Companies — separation of AI Innovation Lab and microbiome partnership

The Estée Lauder Companies (ELC) deliberately separated AI Innovation Lab × microbiome partnership into two distinct tracks during 2024-2025. (i) The AI Innovation Lab with Microsoft (April 2024) — integrating Azure + OpenAI generative AI into brand-facing product development and marketing ^[12]. (ii) A skin-microbiome probiotic research partnership with Nizo (Netherlands) the same year. (iii) An additional 2025 generative-AI partnership with Adobe Firefly for marketing content scaling ^[13].

Strategic background. ELC went through a large business reset in 2024-2025 — a roughly $14B revenue context shaped by China weakness and travel-retail softness, with AI explicitly cited as part of the recovery strategy ^[14]. That background matters — ELC's AI bets are more visible on the marketing and operations side than on the R&D depth side, more so than the other Big-4. The Microsoft partnership is explicitly focused on brand development and marketing workflows ^[12]; Adobe Firefly is explicitly creative production scaling ^[13]. The R&D side is carried by the Nizo microbiome partnership but no peer-reviewed output or disclosed clinical data has resulted from it yet.

M-22 / Estée Lauder Online + AI personalization. ELC integrates an AI personalization layer into its own digital channels (Estée Lauder online, brand sites), but the training data and architecture of that layer are private. Microbiome IP filings root in ELC's post-2002 fermentation library (an in-house yeast / Saccharomyces ferment collection going back to the Aramis era), but no clear disclosure of recent microbiome-AI patent filings is visible at this search horizon (2026-05).

Open question — ELC's microbiome R&D depth. ELC holds 2002+ fermentation-cosmetic heritage (Chapter 1) and microbiome marketing claims under Origins / Estée Lauder / La Mer brands, but zero peer-reviewed outputs at the current search horizon. Whether the Nizo partnership produces clinical output will be the cleanest future signal of ELC's microbiome depth.

10.5 Shiseido — the Voyager platform and the Accenture partnership

Shiseido launched the Voyager AI formulation development platform with Accenture in February 2024 ^[28]. The headline figure — Voyager is trained on over 500,000 data points drawn from Shiseido's century-long R&D archive. In parallel, AI systems assess ingredient biodegradability and safety, tied explicitly to Shiseido's sustainability strategy.

Shape of the data asset. The 500,000 data points are the most explicitly quantified formulation R&D archive among Big-4. The definition of a data point, however, is not disclosed — is it a formulation × stability measurement, a formulation × sensory readout, a formulation × clinical endpoint? The specification is private. This asymmetry matches the broader Big-4 pattern — disclosing the size of a data asset is brag-worthy in marketing, but disclosing its shape would leak IP.

Microbiome R&D on a separate track. Outside Voyager, Shiseido has run Japanese-cohort microbiome R&D, but English-language peer-reviewed outputs are relatively rare (the main channels are Japanese-language in-house publications and JCD conference talks). This mirrors the information asymmetry pattern of the Korean Big-2 (Amorepacific, LG H&H). At CES 2025 Shiseido shared AI-based beauty tech demos alongside L'Oréal and Amorepacific ^[23].

Open question — Voyager's externally auditable output. No publicly disclosed peer-reviewed or clinical-grade output has resulted from Voyager as of this search horizon. Whether Shiseido's output pattern shifts over the next three years from marketing-grade claims (current) to externally auditable evidence (potential) will likely be the most diagnostic signal of where Japanese cosmetic R&D as a whole stands.

10.6 Amorepacific — Vitamin Tree heritage and the disclosure asymmetry (Gap 4)

Amorepacific is Korea's largest cosmetic firm by revenue and operates Sulwhasoo, Innisfree, Laneige, and other global brands. On the AI-microbiome axis, activity is plausibly large but peer-reviewed disclosed output is effectively zero ^[2]. This asymmetry is the cleanest instance of Gap 4 (Korean Big-2 publication invisibility).

Vitamin Tree (Hippophae rhamnoides) ferment heritage. Amorepacific has operated Vitamin Tree (sea buckthorn) strain resources and ferment IP in-house — a heritage line built on Korean plant resources × strain fermentation. In 2024-2025 Amorepacific repackaged this heritage under postbiotic positioning; a postbiotic serum line launched in 2023 self-reports 44% resilience improvement at 28 days ^[1]. That number is from internal clinical work with no peer review and no disclosed protocol.

CES 2025 and the marketing layer. Amorepacific shared beauty-tech demos at CES 2025 ^[23]; its R&D Center page ^[2] names microbiome, fermentation, and AI personalization as in-house R&D tracks. Methodology, datasets, and results are not specified.

KIPRIS patent landscape. The layer English-only readers cannot see is the KIPRIS (Korean patent agency) filings layer. Amorepacific has continuously filed Korean patents related to microbiome × cosmetics (search terms: 마이크로바이옴, 락토바실러스, 발효물 + applicant filter on 아모레퍼시픽) — the R&D depth that does not appear in English press releases is partially visible at the KIPRIS layer. This is exactly the information channel we surface explicitly in Chapter 12 when discussing the Korean R&D blueprint.

Contrast with COSMAX. Within the same Korean industry, COSMAX (a B2B OEM) discloses 5+ peer-reviewed outputs while Amorepacific (a B2C brand owner) discloses near-zero — a direct consequence of disclosure-incentive structure (Gap 4 analysis). For COSMAX, publishing is a sales tool that proves technical credibility to brand customers (the B2C firms themselves); for Amorepacific, publishing risks exposing formulation IP to competitors. By naming this asymmetry explicitly, we prevent Korean R&D planners from misreading no peer review = no R&D.

10.7 LG H&H — the Phyto-Tech heritage and an oral-microbiome MOU

LG Household & Health Care (LG H&H) is Korea's #2 cosmetic firm by revenue, has operated plant-based Phyto-Tech lines for decades, and in 2024-2025 made an explicit move to extend its microbiome R&D from skin into oral. In November 2025 LG H&H signed an MOU with the Sagamsu Medical Foundation (the leading Korean oral-microbiome research entity) at LG Science Park ^[19]. Scope — joint research on preventing oral disease through microbial-environment modulation and mapping oral-systemic health crosstalk (cardiovascular, diabetes).

Why oral. This is the most interesting positioning signal LG H&H sends. Skin microbiome is a competitive space already crowded by Big-4 and Korean OEM players. Oral microbiome is a known proxy for system-level disease (cardiovascular, diabetes, dementia risk markers) and blurs the cosmetic × health boundary more explicitly (cf. Chapter 11's nutraceutical adjacency). Positioning the heritage at the oral × cosmetic × system-health intersection differentiates LG H&H from the Big-4's skin focus.

Disclosure limits. This is MOU-stage and clinical results have not yet appeared. LG H&H's skin microbiome output — like Amorepacific's — is partially visible in the KIPRIS patent layer but barely visible in English-language peer review. Over the next three years, the channel through which LG H&H discloses oral-microbiome results (peer review vs MOU press) will be the most diagnostic signal of Korean Big-2 information-asymmetry trends.

10.8 COSMAX × Dankook FACE-LINK — the Korean OEM running a Big-4-comparable peer-reviewed cohort

COSMAX is Korea's largest cosmetic OEM/ODM and operates the most clearly Big-4-comparable industrial case in this chapter. The core is the FACE-LINK platform of the COSMAX BTI + Dankook + HuNBiome + Yonsei + SNU consortium and the ~1,000-Korean cohort behind it.

FACE-LINK peer-reviewed output. ^[24], published in Frontiers in Cellular and Infection Microbiology, is an integrative classification study combining 16S rRNA microbiome profiling and biophysical skin measurements (TEWL, hydration, elasticity, pigmentation) to classify Korean skin types and aging groups. The company-reported >96% classification accuracy is not externally validated, but the same consortium has commercialized two postbiotic actives, Amioter and Fillerstin ^[8]. This is one of the very few industrial cases in which the closed loop peer-reviewed cohort → commercialized active is publicly operated.

EPI-7 clinical — the second peer-reviewed anchor. ^[18] (Yonsei + COSMAX BTI + Dankook + HuNBiome) is a prospective clinical study evaluating anti-aging endpoints for postbiotics derived from Epidermidibacterium keratini (a novel skin-derived genus discovered by the same Korean consortium). The same Korean consortium also produced ^[17] in Communications Biology (with GIST + Genome and Company), establishing mechanistically that Streptococcus-secreted spermidine drives anti-aging endpoints. The three peer-reviewed anchors — FACE-LINK classification / EPI-7 clinical / Spermidine mechanism — are among the deepest publicly disclosed peer-reviewed microbiome chains produced by any single industrial consortium globally. No Big-4 firm has a publicly disclosed chain of comparable depth.

Second-generation platform and Microbiome AI. COSMAX unveiled a 2nd-Generation Skin Microbiome platform + Microbiome AI in 2024, adding AI-driven postbiotic active screening and personalized recommendation atop the FACE-LINK foundation ^[9]. In 2025 COSMAX announced a partnership with US startup HelloBiome and published joint research using skin-microbiome data from 900 consumers to discriminate healthy vs inflamed skin, producing two postbiotic actives (Amioter, Fillerstin) and a 3-step regimen ^[8].

Bertis proteomics — omics stack extension. COSMAX announced a 2024 partnership with proteomics specialist Bertis to run microbiome × proteome cosmetic lines ^[7]. This extends the FACE-LINK microbiome layer into a protein-expression layer — no Big-4 firm publicly operates an omics integration of this granularity through a single OEM.

Honest position vs Big-4. COSMAX sits below the Big-4 in global revenue but at or above them in the depth of publicly disclosed peer-reviewed microbiome output chain. That is the quantitative basis for the framing "Korea ranks as a global #2-3 region in microbiome cosmetic R&D" (which we develop in Chapter 12). The asymmetry within Korea — COSMAX discloses while Amorepacific and LG H&H do not — is a direct consequence of disclosure-incentive structure, exactly as named above.

10.9 Mid-tier and challengers — Beiersdorf, Galderma, S-Biomedic, Arcaea, Parallel Health, Bertis, HelloBiome

Outside the Big-4 + COSMAX, five to seven specialty challengers hold positions in this space. Each fills a layer the Big-4 do not.

Beiersdorf (parent of Nivea, Eucerin). Beiersdorf acquired a majority stake in Belgian skin-microbiome biotech S-Biomedic in 2022 and integrated it into the in-house Microbiome Design Platform [Beiersdorf, 2022/2024]. The 2024 Capital Markets Day reaffirmed microbiome + skin longevity as a core innovation pillar; in March 2026 Beiersdorf announced a €100M VC fund to accelerate external innovation acquisition and an investment in AI cell-simulation startup Turbine (skin-aging modeling). The 2023 SKINLY study (AI-powered skin research) is the surface signal of Beiersdorf's AI-microbiome integration.

Galderma — the deliberate silence (Gap 10). Galderma is the largest pure-play dermatology company globally, holds Cetaphil and prescription-grade skincare, and shows near-zero public signal on AI + microbiome. The very fact that papers.json has zero Galderma entries among 137 records is itself a data point. Following the March 2024 SIX (Zurich) IPO Galderma is likely in an IR-cautious mode — or has deliberately chosen not to invest at scale here. Absence of disclosure ≠ absence of activity, but absence of disclosure is itself information. Chapter 11 (pharma-derma adjacency) returns to Galderma.

S-Biomedic — RoxP-engineered C. acnes. As Beiersdorf's majority-owned subsidiary, S-Biomedic is developing engineered C. acnes strains that over-express RoxP (radical oxygenase of Propionibacterium acnes) ^[27]. Multiples-over-wild-type RoxP production is reported; clinical data are not public. The cosmetic application of a live engineered strain runs directly into the GMO regulatory vacuum named in Chapter 5 and Chapter 7 (Gap 5).

Arcaea — precision prebiotics. Boston biotech (Ginkgo Bioworks spin-out) developing targeted prebiotic blends for armpit (odor) and scalp (rebalancing) without antimicrobials. Patent WO2024118975A1 ^[3] discloses the mixture claims. An explicit strategic choice to sidestep the live-engineered regulatory vacuum by moving to the prebiotic side.

Parallel Health — consumer phage cocktail. First consumer-facing skin-phage product line, supplying an at-home skin-microbiome test → curated phage cocktail ^[25]. Targets disease-associated C. acnes / S. aureus strains while sparing beneficial commensals. External validation is absent; customer testimonials are the main publicly available data.

HelloBiome. US startup whose most visible output is the COSMAX partnership ^[8]. Provides a SaaS-like layer that maps microbiome profile → formulation strategy; the R&D depth leans on COSMAX's OEM infrastructure.

Bertis — proteomics partner. Korean proteomics company whose surface signal is the COSMAX anti-aging proteome cosmetics collaboration ^[7]. Fills the adjacent omics layer of the microbiome stack.

Eligo Bioscience — CRISPR-armed phage. Paris-based; develops engineered phages delivering CRISPR-Cas payloads to selectively kill disease-associated C. acnes strains while sparing beneficial commensals, using the same technology stack as Brödel 2024 Nature in situ base editing ^[11]. Clinical-stage data are not public; in ^[4]'s review it carries the highest technology ceiling in the live-engineered category.

10.10 The Korean information asymmetry (Gap 4) — the KIPRIS channel and the limit of English-only reading

Korean cosmetic industry microbiome R&D activity is distributed across multi-layer channels that an English-only reader cannot see directly — the KIPRIS patent database, KSCS society talks, domestic trade press (코스인 / 화장품신문 / 뷰티누리), and internal R&D bulletins. Naming this asymmetry explicitly at the end of this chapter is the most practical takeaway for Korean R&D planners.

KIPRIS searchability. Amorepacific and LG H&H likely surface microbiome R&D output in Korean patents that does not appear in English press releases. Search terms: 마이크로바이옴, 락토바실러스, 발효물, 포스트바이오틱, 미생물 군집, 균총 + applicant filter. The KIPRIS English interface is partial, but applicant + IPC classification filtering with chronological ordering works.

Information-channel asymmetry — Big-4 vs Korean OEM. Big-4 firms center on global PR stages (VivaTech, SXSW, CES, R&I editorials). The Korean Big-2 center on KSCS, domestic trade press, and KIPRIS. COSMAX is visible on both — peer-reviewed plus global media — which is exactly why it is almost the only Korean firm with Big-4-comparable external auditability.

Frame for decision-makers. A Korean R&D planner who wants to know what AI-microbiome stack Amorepacific or LG H&H actually operates should not conclude from English search alone. The minimum protocol is three-source triangulation — KIPRIS patent layer + KSCS conference talks + domestic trade press. This chapter's company comparison reflects only the English-channel depth, so the actual R&D depth of Korean firms is larger than this chapter shows; how much larger is, by the nature of the asymmetry, not externally calibratable.

10.11 Cross-firm comparison table — multi-axis frame

The table below sorts the firms in this chapter on a five-axis frame. Cell judgments reflect publicly disclosed material as of May 2026 and do not calibrate against actual internal R&D depth.

How to read the table. Once again — disclosed depth and actual depth are not the same. An empty Galderma cell is not "no activity"; it is "not disclosed". The zero peer-reviewed count for the Korean Big-2 is not "no R&D"; it is "the English-channel route is not used". Read this as a table of externally visible industry signal, not a table of internal industrial-activity ranking.

10.12 Open Questions

1. Big-4 vs Korean OEM advantage path. Big-4 dominates in global revenue, but the COSMAX × Dankook consortium is comparable in disclosed peer-reviewed output chain. Over the next five years, where will the first peer-reviewed clinical result for an AI-designed cosmetic active appear? That is the cosmetic-industry equivalent of Gap 1. One reading of the ^[16] PRISMA distribution suggests Korean OEMs are likelier than French / US / Japanese majors.

1. Will the Korean Big-2 switch disclosure channels? Amorepacific and LG H&H face an incentive trade-off between global competitive pressure and Korean market-share defense when considering English-language peer review. Over the next three years, the migration pattern from KSCS conference talks to English-language Frontiers / International Journal of Cosmetic Science publications is the most visible signal of where the information asymmetry is heading (Gap 4 follow-up).

1. The end of Galderma's deliberate silence. Galderma is likely to disclose an AI-microbiome program by 2027 — if it does, the reading is that the silence was IR caution; if it does not, the reading is that it has chosen not to bet at scale. Either way it is information.

1. First regulatory rulings on consumer phages and engineered strains. Which firm among S-Biomedic, Eligo, Parallel, Arcaea triggers the first major regulatory decision (FDA OTC monograph supplementation / EU CTR live-cosmetic guideline / Korean MFDS functional-cosmetic category extension)? Chapter 7's Gap 5 narrows to a Chapter 10 commercial event.

1. COSMAX's global OEM expansion. How far COSMAX leverages its peer-reviewed chain in sales to English-speaking brand customers is the best leading indicator of whether Korea becomes the default for global microbiome OEM. Watch COSMAX's US / EU brand-customer expansion from 2026 to 2028.

1. Shiseido Voyager's externally auditable output. Does Voyager produce a peer-reviewed or clinical-stage output by 2027? This is the cleanest signal of where Japanese major firms' information-asymmetry pattern (similar to Korean Big-2) stands.

(Chapter 4) FACE-LINK is referenced again as cosmetics' first publicly disclosed AI-strain-screening cohort. (Chapter 8) Unilever × IBM's Carrieri 2021 reappears as the methodological backbone of industrial AI formulation. (Chapter 11) The next chapter extends the Big-4 / Korean / challenger industry frame into the adjacent template of pharma and nutraceutical industries. (Chapter 12) The three structural gaps — information asymmetry, data-asset gating, and regulatory vacuum — integrate into a blueprint for new research planning.

References

1. Amorepacific Group (2024). Amorepacific postbiotic serum line and microbiome positioning 2024-2025. Korean cosmetic press 2024-2025; CES 2025 Beauty Tech coverage. [Amorepacific Group, 2024]
2. Amorepacific R&D Center (2024). Amorepacific R&D microbiome research output — Vitamin Tree ferment + postbiotic strategy. Amorepacific in-house publications + KSCS 2024-2025. [Amorepacific R&D, 2024]
3. Arcaea (2024). Arcaea precision prebiotics for armpit and scalp microbiome modulation. Arcaea patent WO2024118975A1; HPC Today feature 2025. [Arcaea, 2024]
4. Atallah, C., El Abiad, A., El Abiad, M. et al. (2025). Bioengineered Skin Microbiome: The Next Frontier in Personalized Cosmetics. Cosmetics 12(5):205. [Atallah et al., 2025]
5. Beiersdorf AG (2022/2024). Beiersdorf majority stake in S-Biomedic — Microbiome Design Platform integration. Beiersdorf press 2022 + 2024 Capital Markets Day. [Beiersdorf, 2022/2024]
6. Carrieri, A. P., Haiminen, N., Maudsley-Barton, S. et al. (2021). Explainable AI reveals changes in skin microbiome composition linked to phenotypic differences. Scientific Reports 11:4565. [Carrieri et al., 2021]
7. COSMAX × Bertis (2024). COSMAX and Bertis join forces to redefine anti-aging skincare with skin proteome cosmetics. The Monodist, 2024-2025. [COSMAX × Bertis, 2024]
8. COSMAX × HelloBiome (2025). Korean beauty manufacturer COSMAX × HelloBiome microbiome-powered personalized care. WWD / Personal Care Insights, 2025. [COSMAX × HelloBiome, 2025]
9. COSMAX USA (2024). COSMAX unveils 2nd-Generation Skin Microbiome platform + Microbiome AI. Global Cosmetic Industry, 2024. [COSMAX USA, 2024]
10. Di Guardo, A., Trovato, F., Cantisani, C. et al. (2025). Artificial Intelligence in Cosmetic Formulation: Predictive Modeling for Safety, Tolerability, and Regulatory Perspectives. Cosmetics 12(4):157. [Di Guardo et al., 2025]
11. Eligo Bioscience (2024). Eligo Bioscience CRISPR-Cas-armed phage for C. acnes. Eligo patent EP4240843A1 + Atallah 2025 review citation. [Eligo, 2024]
12. The Estée Lauder Companies + Microsoft (2024). ELC × Microsoft AI Innovation Lab announcement. ELC press release, Apr 2024. [ELC + Microsoft, 2024]
13. The Estée Lauder Companies + Adobe (2025). ELC × Adobe Firefly generative AI for marketing production scaling. Adobe news, Mar 2025. [ELC + Adobe, 2025]
14. The Estée Lauder Companies (2025). Inside Estée Lauder's $14B reset — AI, brand trouble, and a travel retail retreat. Glossy, Mar 2025. [Glossy Inside ELC, 2025]
15. Gueniche, A., Perin, O., Bouslimani, A. et al. (2022). Advances in Microbiome-Derived Solutions and Methodologies Are Founding a New Era in Skin Health and Care. Pathogens 11(2):121. [Gueniche et al., 2022]
16. Haykal, D., Flament, F., Amar, D. et al. (2025). Cosmetogenomics unveiled: a systematic review of AI, genomics, and the future of personalized skincare. Frontiers in Artificial Intelligence 8:1660356. [Haykal et al., 2025]
17. Kim, G., Kim, M., Kim, M. et al. (2021). Spermidine-induced recovery of human dermal structure and barrier function by skin microbiome. Communications Biology 4:231. [Kim et al., 2021]
18. Kim, J., Lee, Y. I., Mun, S. et al. (2023). Efficacy and Safety of Epidermidibacterium Keratini EPI-7 Derived Postbiotics in Skin Aging: A Prospective Clinical Study. International Journal of Molecular Sciences 24(5):4634. [Kim et al., 2023]
19. LG Household & Health Care + Sagamsu Medical Foundation (2025). LG H&H × Sagamsu Medical Foundation oral microbiome MOU. Etoday / EBN / The Public, Nov 2025. [LG H&H + Sagamsu, 2025]
20. L'Oréal R&I Press (2023). L'Oréal acquires Lactobio to strengthen microbiome research. L'Oréal press, Dec 2023 (closing Q1 2024). [L'Oréal, 2023]
21. L'Oréal R&I (2024). L'Oréal Beauty Tech leadership at VivaTech 2024 — AI Skin Genius, Modiface, microbiome direction. L'Oréal press, May 2024. [L'Oréal, 2024]
22. L'Oréal R&I (2024). The Future of Cosmetics Is Playing Out In The Microbiome. L'Oréal R&I editorial, 2024. [L'Oréal R&I, 2024]
23. L'Oréal Modjoul Beauty Tech (2025). Amorepacific · Shiseido · L'Oréal — 5 Beauty Tech Innovations Not to Miss at CES 2025. Formes de Luxe, CES 2025 coverage. [L'Oréal Modjoul, 2025]
24. Mun, S., Jo, H., Heo, Y. M. et al. (2025). Skin microbiome-biophysical association: a first integrative approach to classifying Korean skin types and aging groups. Frontiers in Cellular and Infection Microbiology 15:1561590. [Mun et al., 2025]
25. Parallel Health (2024). Parallel Health phage cocktail for skin inflammaging and redness. Parallel Health product page + HPC Today 2025. [Parallel Health, 2024]
26. POND'S (Unilever) (2024). POND's Skin Institute microbiome analyzer — 60-minute in-store consumer device. Unilever press, May 2024. [POND'S, 2024]
27. S-Biomedic (2024). S-Biomedic engineered C. acnes RoxP-expressing probiotic strain. Atallah 2025 review citation + S-Biomedic press / HPC Today 2025. [S-Biomedic, 2024]
28. Shiseido + Accenture (2024). Shiseido develops AI systems for ingredient biodegradability and Voyager formulation platform. Global Cosmetics News, Feb 2024. [Shiseido + Accenture, 2024]
29. Unilever Beauty & Wellbeing R&D (2025). How Unilever's pioneering skin microbiome research is shaping product innovation. Unilever news, 2025. [Unilever, 2025]
30. Unilever Beauty & Wellbeing (2025). SXSW 2025 AI/ML/data behind Unilever's latest launches. Unilever news, Mar 2025. [Unilever, 2025-SXSW]
31. Unilever (2026). How AI is transforming innovation in Unilever Beauty & Wellbeing. Unilever news, 2026. [Unilever, 2026]
32. Unilever Discover Science (2025). Unilever scientists discover link between skin microbiome and mental wellbeing. Unilever USA press, 2025. [Unilever Discover Science, 2025]