# Into: Yanez MIID Miners generate adversarial identity data that compliance teams cannot legally produce themselves. Banks must test KYC and sanctions screening against realistic name variations, transliterations, and now deepfake face images. Real customer data is off limits. Yanez MIID generates the test set. // The adversarial dataset for KYC. --- > New to Bittensor? Start here. Experienced users can skip to the full analysis. ### What is Yanez MIID? Yanez MIID (Multimodal Inorganic Identity Dataset) is subnet 54. It produces synthetic identity variations that financial institutions use to stress test fraud detection, sanctions screening, and KYC pipelines. **The simple version:** Sanctioned individuals use small spelling tweaks, alternate transliterations, and lookalike documents to slip past screening systems. Banks need a constantly evolving library of those evasion tactics to test against. Miners on SN54 produce them on demand. **Centralized equivalent:** Hazy, Mostly AI, and Gretel.ai produce synthetic data for analytics. MIID is narrower and meaner, focused specifically on identity evasion patterns for AML, sanctions, and IDV testing. **How it works:** - **Miners** receive identity challenges from validators and return KAV variations (name, date of birth, address) plus face image variations from validator seed images. 198 active miners currently compete on the network. - **Validators** issue the challenges, score outputs on accuracy, novelty, constraint adherence, and adversarial value, then update miner reputation across cycles. --- ### Why This Matters - **The problem it solves:** AML, sanctions screening, and KYC systems are only as good as the adversarial cases they have been tested against. Privacy law blocks the use of real customer records for testing. Static synthetic datasets go stale the moment threat actors adapt. - **The opportunity:** Every regulated financial institution has a mandatory model validation cycle for screening systems. Compliance is a non-discretionary spend, and the dataset needs to keep evolving with the threat landscape. - **The Bittensor advantage:** Decentralized competition produces a wider distribution of evasion patterns than any single in-house red team. Miners are paid to be creative about how a fraudster might mangle a name or recapture a passport photo. - **Traction signals:** 198 active miners, 6 GitHub contributors, with the most recent commit on 2026-05-20. Phase 4 (face-based adversarial testing for KYC) is now live in the repo, on schedule with the published Q1 2026 roadmap. --- ## Full Analysis **Category:** Deepfake Detection and Security | **Centralized Competitor:** Hazy, Mostly AI, Gretel.ai, Tonic.ai Yanez MIID sits at an unusual crossroads. The buyer profile is enterprise compliance, which is one of the slowest sales cycles in software, but the demand driver is regulatory, which means budget exists by default. Most Bittensor subnets chase consumer or developer surface area. MIID is positioned for procurement officers. **Mechanism:** Validators send mixed identity challenges to miners. KAV requests cover name, date of birth, and address variations, including phonetic, orthographic, rule-based, and transliteration-based alterations. Image challenges send seed face images that miners transform along axes such as pose, lighting, expression, and background. Scoring is multi-dimensional: accuracy against the challenge constraints, novelty versus prior submissions, and how realistically adversarial the output is for downstream model testing. Validators run online validation for immediate weight setting where applicable, then post-validate for novelty and quality, carrying that score into the next cycle. The repo shows the team executing on the published roadmap. Recent commits from Asem Othman and Omar A. Abaza, the two highest-volume contributors, focus on the face variation pipeline: image encryption, ada_face fixes, variation method additions, and the reward allocation backlog. Phase 4 was scheduled for Q1 2026 and the code is in the repo now. The contributor distribution is healthy: 6 active contributors with the top four each above 70 commits. Market state is sober. Price sits at 0.00610 TAO with a market cap of 29,873 TAO. Emission share is 0% on the latest snapshot with miner burn at 65.16%. Root in pool is 9,580.80 TAO. Despite the emission state, flows are slightly positive: net 24h inflow of 280.82 TAO and 7d net inflow of 114.94 TAO. Buy volume over the past day was 699.78 TAO against 418.96 TAO sold. 24h price is up 6.22%, the 7d is down 4.08%, and the 30d is down 8.87%. Top 100 Gini sits at 0.641 with an HHI of 0.047. The disconnect between active development and emission is the story. The team is shipping Phase 4 deepfake work into a market that is currently routing emission elsewhere. If the face-image pipeline starts attracting validator weight, the emission picture changes. If it does not, the subnet is leaning on its enterprise pipeline at yanez.ai to monetize the dataset directly. --- ### Risk Factors - **Zero current emission:** Emission share is 0% with 65.16% miner burn on the latest snapshot. The subnet must earn back validator weight to capture meaningful inflation, or rely on off-chain commercial revenue. - **Enterprise sales cycle:** Compliance procurement is slow. The yanez.ai commercial surface needs paid pilots to translate codebase activity into a sustainable token thesis. - **Niche dataset competition:** Centralized synthetic data vendors are well-funded and integrated into enterprise procurement. MIID has to win on quality and adversarial breadth, not on integration polish. - **Roadmap execution risk:** Phases 5 through 11 promise biometric attack families, synthetic documents, financial transaction modeling, and 3D identity avatars through 2027. The current team must keep shipping at this cadence to land the full vision. --- Another subnet, unpacked.