Why Yubo Is Rebuilding for Adults
On November 1, 2025, Yubo officially became an 18+ platform — a structural acknowledgment that the user base the company designed for had become a minority on its own service. For more than a year before the policy change, users 18 and older comprised the majority of Yubo's global community, which numbered more than 80 million across over 140 countries as of early 2024. In an in-app survey, 92 percent of respondents voted for an adult-only space. The teen segment had dwindled to under 5 percent of total accounts.
Founded in 2015 by CEO and co-founder Sacha Lazimi, Yubo launched with a specific thesis: teens needed a place to meet new people safely, without the toxicity and performative pressure of mainstream social media. The app introduced live video in 2018, removed likes and follows, and built a "social discovery" model around real-time interaction rather than curated feeds. By 2022, TechCrunch reported 60 million sign-ups, nearly all of them Gen Z users aged 13 to 25. Revenue from à la carte premium features and subscriptions had grown from €7 million in 2019 to €25 million, TechCrunch's data shows, with no advertising.
The cohort that joined at 13 in 2015 turned 23 in 2025. The platform's core demographic aged alongside the product, and the safety architecture built to separate minors from adults (age-gated communities, facial age estimation at sign-up, 24/7 human moderation of livestreams, behavioral monitoring for age mismatch) became a system optimized for a shrinking edge case.
"Since our launch 10 years ago, Yubo's core demographic has grown up – from young teens to young adults — making it imperative that we, as a platform, 'grow up,' too," Lazimi said in the November announcement. The company framed the move as positioning Yubo to "serve a single audience with clarity and purpose."
That clarity has engineering consequences. Every product decision, including live-streaming room dynamics, swipe mechanics, chat architecture, profile customization, and geolocation features, was previously constrained by the need to maintain compliant, separable experiences for minors and adults. The trust-and-safety operation, spanning 24 countries and five continents under operations heads Alexander Boyd and Katie Boyd at the Jacksonville, Florida headquarters, runs a three-tier escalation model: routine moderation at the base, market-specific legal requests in the middle, and a dedicated Law Enforcement & Escalation Coordinator at the top. A Safety Board, including CEO Lazimi, COO Marc-Antoine Durand, Head of Legal & Public Policy Sharone Franco, Head of Trust & Safety Operations Margaux Liquard, and independent advisors from NCMEC, Thorn, The Diana Award, and law enforcement, sets policy and reviews edge cases.
Removing the minor segment does not simplify the safety stack. It inverts it. The age-verification funnel that processed over 290 million checks in 2025 must now prove a negative at scale: that no minors slip through into an adult-only environment. The moderation models trained to flag teen-risk patterns (grooming, CSAM, age-inappropriate contact) must be retrained for adult-risk patterns (financial fraud, romance scams, identity theft, non-consensual intimate imagery). The Global Anti-Scam Alliance partnership announced in 2025 signals the shift.
The product roadmap reflects the inversion. Upcoming updates focus on expanded profile customization, granular control over group livestream interactions, and optional geolocation for local connection — features that make sense for young adults navigating relocation, dating, and interest-based community. The engineering trade-off is no longer "how do we protect minors on an adult internet" but "how do we build the best adult social discovery product on top of a safety foundation that was over-engineered for a different era."
Yubo's bet is that the safety-by-design infrastructure it pioneered (the first platform-wide age assurance in 2022, real-time video and audio moderation, DSA-aligned transparency reporting) becomes a competitive moat in a social discovery market. Verified Market Research valued the market at $3.47 billion in 2025, projecting 11.3% CAGR through 2033. PMarketResearch placed it at $14.98 billion in 2025 with 8.52% CAGR through 2032. TikTok optimizes for passive scrolling. Snapchat optimizes for ephemeral messaging. Yubo's architecture, built for live, uncurated interaction among strangers who verify their identity, now serves a single, verified adult audience.
The AI Stack That Verifies Ages at Scale
Yubo did not wait for regulators to force its hand. In May 2022, the company began rolling out facial age estimation across its entire user base — a move that made it the first major social platform to require age verification for 100 percent of users. The system went live on iOS first; Android, representing roughly one-tenth of the base, followed within weeks. By year end, the mandate was complete.
The engine behind that mandate is Yoti, a UK-based digital identity company whose facial age estimation model analyzes a selfie and outputs an age range — no document upload, no government ID, no persistent biometric storage. The model was trained on millions of labeled images with explicit consent, and Yoti says it can estimate ages for 6- to 12-year-olds within 1.3 years and 13- to 19-year-olds within 1.5 years. On the aggregate data set the company publishes, the system reaches 98.9 percent accuracy at distinguishing adults from minors.
At Yubo's scale, those numbers translate to operational throughput. The platform processed an average of 6,500 verifications per day in 2021, before the full rollout. By December 2024, Yoti's case study recorded more than 290 million age checks completed on Yubo. About 87 percent of users pass on the first attempt. The rest are routed to a secondary flow (document check or human review) before they can access livestreams or swipe features. In 2021 alone, the discrepancy mechanism caught roughly 67,000 accounts per month.
The architecture is multi-layered. Facial estimation sits at the front gate. Behind it, a rules engine enforces age-gating: users estimated under 18 are placed in a separate community pool, unable to join adult livestreams or match with adult profiles. Accounts flagged for discrepancy, where the estimated age conflicts with the declared birthdate, are suspended automatically.
Why build this before the 18+ pivot? Because the pivot was not a product decision — it was a demographic acknowledgment. By late 2024, roughly 70 percent of Yubo's users were already 18 or older. The age-verification stack, completed in 2022, gave the company ground truth on that shift. Without it, Yubo would have been guessing at its own composition. The same infrastructure that once protected minors from adults now protects adults from minors, and both from bots. In an October 2024 survey of over 10,000 users, nearly half reported fewer bot encounters since verification launched; 53 percent said it increased trust that other users are real people.
The stack also bought regulatory runway. The UK Online Safety Act took effect in 2023. Australia banned under-16 social media use in 2024. California, Utah, Texas, and other U.S. states have passed or are debating age-assurance laws. Yubo's system predates all of them.
Every product decision now routes through that verification layer. Livestream matching, swipe algorithms, geolocation features, and upcoming profile customization tools all consume the same age signal. The engineering trade-off is latency: the selfie check adds seconds to onboarding, and the fallback flows add friction for edge cases. Yubo accepts that friction. As Lazimi put it in 2022, the company is "willing to drive away a portion of new users in order to prioritize making Yubo as safe as it can possibly be." The revenue model (subscriptions and in-app purchases, not advertising) makes that calculus viable.
The result is a rare thing in consumer social: a safety system that is not a bolt-on. It is the substrate.
How Safety Systems Change When the Minor Goes Away
The safety architecture Yubo built over a decade was designed for a different threat model. When the platform launched in 2015, the core risk was adults preying on minors — grooming, CSAM, age-inappropriate contact. The response was a "teen shield" stack: strict age gates, real-time video moderation on every livestream, and a Trust & Safety operation trained to spot child-exploitation signals. After rolling out verification, Yubo saw a 20 percent drop in user count but stronger engagement, more authentic community, and sustainable trust-based growth. Seventy-nine percent of users said they felt safer; 60 percent said age verification improved their experience. Bots and scams decreased. But the shield was built for a population that no longer exists.
The pivot to 18+ doesn't remove safety work — it inverts it. The new "adult guardrails" must address harms that matter to young adults: harassment, hate speech, non-consensual intimate imagery, financial scams, coordinated inauthentic behavior, and violence. Child-safety hash databases (NCMEC, IWF, PHAROS) remain active; Yubo still escalates to those bodies, but they're no longer the primary detection target.
Real-time video moderation stays, but the policy layer changes. Safety Specialists monitoring 24/7 across 24 countries and five continents now apply adult-content policies. The behavioral prompts (those in-the-moment nudges developed with behavioral experts) shift from teen-focused warnings to adult-harm interventions. The human-in-the-loop pipeline reweights toward nuanced harassment adjudication and scam-cluster investigations.
The operational structure reflects this. Liquard's team separates into distinct pods: a Policy team writing adult-community guidelines, a Program team building detection workflows for adult harms, an Operations team staffing the 24/7 review queues with specialists trained on adult-content nuance. "You also need to have clear people dedicated to clear things," Liquard said in a 2024 interview. "A policy team, a program team, an operation team that can advocate at the right level of a product stage or project implementation their part of the story and why it's important."
Regulatory exposure shifts too. COPPA and GDPR-K compliance, built for minors, becomes secondary. The EU Digital Services Act, UK Online Safety Act, and emerging U.S. state laws for adult-platform accountability take center stage. Yubo's AFNOR Spec, co-drafted with Meta, Yoti, Arcom, and CNIL, already structures its safety operations around three pillars: account verification and age assurance; content moderation and risk detection; awareness and transparency. The 18+ transition moves weight from pillar one (now a gate, not a separator) to pillar two.
The engineering implication: every model, every queue, every escalation path needs retraining or rebuild. The age-estimation model at signup still runs; now it's a hard 18+ gate with ID fallback, but the downstream classifiers swap target labels. The 290 million age checks conducted to date become the training corpus for a new problem: not "is this user a minor?" but "is this adult user behaving in ways that violate adult-community standards?"
Inside the 24/7 Human Layer
Yubo's trust-and-safety operation runs on the three-tier model described in its safety documentation. The Jacksonville office, opened in 2020, anchors the Americas expansion; the company plans to grow the U.S. team to roughly 10 people across safety, marketing, and communications. Before Jacksonville, moderation was coordinated from Paris with remote specialists. The shift to a physical U.S. hub reflects the platform's user geography: TechCrunch's figures put U.S. live-stream participation growth at 350 percent from March 2020, Latin American participation 1,200 percent from June 2020, according to TechCrunch, and the need for time-zone coverage that matches peak live traffic.
Moderators work 24/7. During live sessions, AI takes second-by-second screenshots and flags nudity, suggestive content, drugs, weapons, blood, and violence to human reviewers in real time. Yubo was the first major social platform to deploy real-time video moderation for livestreams. Every flagged moment passes through a human before enforcement action — a design choice that keeps false positives from silencing legitimate conversation. The same human-in-the-loop requirement applies to child-safety hash matches: the company says it always pairs database hits with human review before escalating to NCMEC, IWF, or PHAROS in France.
The 18+ transition changes the workload in two directions. First, the volume of age-discrepancy suspensions should drop sharply now that the minimum age is 18 and the platform no longer onboards 13-to-17-year-olds. Second, the nature of escalations shifts. Teen-focused harms (grooming, CSAM, underage access) remain a legal obligation, but the day-to-day queue increasingly surfaces adult-to-adult disputes: harassment, non-consensual intimate imagery, hate speech, and fraud. The three-tier model absorbs that shift by routing routine adult-content moderation to tier one, market-specific legal requests (GDPR, DSA, state-level U.S. laws) to tier two, and imminent-threat or law-enforcement referrals to the Escalation Coordinator.
Staffing follows the queue. Job postings for remote Trust & Safety Specialists emphasize privacy compliance, child-safety law, and "objective enforcement decisions" — language that signals the dual mandate: protect the residual minor population that may still attempt access, while applying adult-community standards to the 90-percent-plus user base that is now 18 or older. The Jacksonville hires sit alongside the existing global network, not replace it; the company's safety page describes the team as "a global network of specialists working around the clock."
Partnerships extend the human layer beyond Yubo's payroll. The NCMEC reporting tool integrated into Yubo's in-app reporting flow lets any user flag a suspicious profile directly to the center. Membership in the Technology Coalition's Project Protect, the Internet Watch Foundation, and the Global Internet Forum to Counter Terrorism adds shared hash databases and cross-platform signal sharing.
The engineering implication is clear: every AI model that scores a live frame or a chat message must output a confidence interval that a human can act on in seconds. The 18+ pivot does not reduce the need for human review; it reweights the categories those humans judge.
What Adult Social Discovery Demands
Yubo's product roadmap has always been constrained by a split audience. For years the app served 13-year-olds and 25-year-olds on the same infrastructure — two cohorts that, as the company's blog puts it, "often live in completely different, hard-to-pair worlds." The 18+ pivot removes that constraint. Product updates announced alongside the transition focus on three levers: expanded profile customization, finer-grained control over group livestreams, and optional geolocation for users who want local connections. Each lever touches a core feature (live streaming, swiping, chat, or community tools) and each forces a distinct engineering trade-off.
Live streaming: from guardian rails to adult guardrails
Yubo's signature feature is the group livestream room. Unlike Twitch or TikTok Live, there are no creators broadcasting to fans; every participant is on equal footing, playing games or talking in real time. That design choice ("no plans to move in that direction," Lazimi told TechCrunch in 2022) means the moderation surface is combinatorial: every additional user in a room multiplies the pairwise interactions the safety stack must monitor. When the user base was majority-minor, the system defaulted to the strictest tier: second-by-second screenshot analysis flagging nudity, partial nudity, and those same categories to human moderators in real time.
With minors off the platform, the engineering team can relax content thresholds, but the scale problem intensifies. Adding richer, longer adult sessions means more video minutes per daily active user, more concurrent rooms, and a higher ceiling for simultaneous moderation queues.
Swiping and discovery: geolocation as a first-class signal
The swipe mechanic that seeds rooms is also being rewired. The PRNewswire announcement notes "optional geolocation settings for those who want to prioritize local connections." For a teen app, precise location was a liability. For young adults, proximity is a feature: meeting someone in the same city or university town converts online chat to offline plans. Engineering-wise, this means promoting geolocation from a permissioned, opt-in field to a core ranking signal in the discovery algorithm, while still honoring the "optional" contract for users who disable location.
Chat: encryption, ephemerality, and the fraud vector
One-to-one and group chat inherit the same shift. Adult users expect disappearing messages, media auto-expiry, and increasingly end-to-end encryption for sensitive conversations. Yubo's current architecture routes all messages through the moderation pipeline before delivery. The compromise under discussion is a hybrid approach that preserves proactive detection (87 percent of alerts already come from automated signals, not user reports) while introducing a key-management surface the platform never needed when every message was plaintext. The fraud vector also changes. That partnership flags romance scams, crypto pitches, and money-mule recruitment as top adult risks.
Community tools: profiles, groups, and the identity stack
Expanded profile customization sounds cosmetic; underneath it is an identity-stack rewrite. Teen profiles were minimal because rich self-disclosure increased grooming risk. Adult profiles can support verified badges, linked social handles, pronoun fields, and free-form bios. Each field is a new PII column, a new consent surface for GDPR/CCPA, and a new vector for doxxing or impersonation. The engineering response is a privacy-by-design schema with default-off visibility toggles, audit logs, and deletion cascades tested against large-scale synthetic datasets.
The meta-trade-off: safety velocity vs. product velocity
All four workstreams share a bottleneck: the safety-by-design review gate. Every product spec still passes through Liquard's Trust & Safety team, the Safety Engineering squad, and the external Safety Board. The 18+ policy doesn't remove the gate; it changes the threat model. Lazimi's framing ("we are not waiting for regulation to do safety features") means the gate stays. The only lever is tooling: the Safety Engineering team is building the adult rule set as a parallel policy-as-code pipeline so that a product manager can get a safety signal in the IDE instead of a multi-week legal review. That pipeline is the real product rebuild — not the UI, not the algorithm, but the ability to ship adult features at adult speed without regressing the safety floor that made the pivot possible.
Can Live Rooms Outlast the Algorithmic Feed?
The social discovery software market reached $3.47 billion (Verified Market Research) to $14.98 billion (PMarketResearch) in 2025, with projections clustering around 8–13 percent annual growth through the early 2030s. Dating-focused discovery commands roughly 62 percent of that market (PMarketResearch). Mobile applications represent 87 percent of deployment (PMarketResearch). Yubo's more-than-80-million-user base and roughly €25 million in annual revenue, according to TechCrunch, place it in the top tier of a fragmented field, but the competitive pressure comes less from niche rivals than from the hyperscale platforms rewriting the rules of discovery itself.
TikTok's algorithmic feed has turned non-follower content into the default consumption mode. Non-follower views rose from 31 percent in 2023 to nearly 60 percent by late 2025; brand video views on the platform jumped 44 percent in the same window. Instagram Reels followed a similar trajectory, non-follower share climbing from 30 percent to almost half, views up 20 percent. Both platforms now optimize for entertainment score — content scoring above a five-out-of-ten threshold holds viewers for nine seconds versus six for lower-scoring material. The implication for Yubo is structural: the "social" in social discovery is being unbundled from the "follow graph" and reattached to an interest graph that TikTok and Meta own at billion-user scale. Yubo's live, real-time interaction model (group video rooms, swipe-to-match, concurrent chat) is a deliberate architectural counterbet. It cannot out-algorithm TikTok. It has to out-participate it.
Snapchat's historical strength in ephemeral, camera-first communication keeps it adjacent to Yubo's core use case. The Digital Markets Act in the EU has opened alternative app stores (Epic Games, Aptoide, Onside) and early listings include social discovery apps like PeopleDrop. That distribution shift lowers the moat for any app that can acquire users outside the Apple/Google duopoly, but it also floods the top of the funnel with low-retention experiments.
The funded competitor set is consolidating around AI-enhanced matching. Bumble launched Bumble for Friends in 2025 with group matching. Match Group added AI-powered Notes to Hinge to seed conversation prompts. The Meet Group partnered with Twitch to embed live-streaming discovery inside Skout. OpenAI's Sora app, slated for a TikTok-style algorithmic feed of generated video, enters as a wildcard: generative content as social lubricant. Threads hit 115 million daily actives in mid-2025, up 128 percent year-over-year, while X declined 15 percent to 132 million. Bluesky captured the political-exodus cohort. None of these are live-social-discovery products, but they compete for the same attention budget — Deloitte's 2025 survey puts average U.S. media time at six hours daily, with Gen Z spending 54 percent more of it on social and user-generated content than the average consumer.
BeReal, at 40 million monthly users (85 percent Gen Z), began rolling out in-feed ads and full-day brand takeovers in the U.S. in April 2025, hiring former TikTok executive Ben Moore to lead commercialization. Its dual-camera authenticity pitch overlaps Yubo's "real-time, not curated" positioning, but BeReal remains asynchronous, one-post-per-day. Yubo's synchronous rooms (concurrent live streams, persistent chat, moderation tooling built for 24/7 operation across 24 countries) are a different engineering beast entirely.
Yubo's 18+ pivot sharpens its differentiation. The platform no longer straddles teen safety and adult expression; it can build for adult social discovery (dating-adjacent but not dating-app, creator-economy-adjacent but not creator-platform) without the regulatory overhang that constrains TikTok, Meta, and Snap. The engineering priority shifts from age-gating infrastructure (largely solved) to adult-specific trust-and-safety: behavioral models for harassment, fraud, and spam that differ materially from minor-protection classifiers; monetization primitives (gifting, subscriptions, virtual goods) that TikTok Live and Twitch have proven at scale; and recommendation systems that optimize for conversation depth rather than watch time.
The market's growth trajectory (Asia-Pacific leading on smartphone penetration and cultural shift toward digitized social discovery in India and China) suggests the next wave of users for Yubo may not come from its current Europe/NA/LATAM strongholds. The Jacksonville HQ opening signals U.S./LATAM intent. But the competitive moat will not be user count. It will be whether Yubo's live, moderated, adult-only rooms can sustain a retention curve that algorithmic feeds (infinite, passive, ad-optimized) structurally undermine. The guardrails that once kept teenagers in now keep the algorithm out. The rest is product velocity.
Working in AI? Zero G Talent tracks the openings: see every open Databricks role, browse AI jobs, openings at Anthropic, and the people building the field.