With the amount of new subnets being added it can be hard to get up to date information across all subnets, so data may be slightly out of date from time to time
With the amount of new subnets being added it can be hard to get up to date information across all subnets, so data may be slightly out of date from time to time
Bittensor Subnet 54 (MIID) is a specialized decentralized network designed by Yanez Compliance to generate synthetic “inorganic” identities for testing and strengthening financial crime prevention systems. In essence, MIID serves as a data-generation engine that creates realistic, varied identity profiles (names, documents, biometrics, etc.) to stress-test compliance tools (e.g. sanctions screening, KYC verification, fraud detection) under adversarial conditions. Unlike many experimental AI subnets, MIID is backed by an established compliance technology company with real enterprise use-cases, ensuring the project is grounded in practical demand and not just speculation. Below is a comprehensive breakdown of MIID’s purpose, operation, architecture, team, and roadmap.
MIID’s core purpose is to produce high-quality synthetic identity data to improve and validate financial crime defense systems. Financial institutions face ever-evolving evasion tactics – from fake identities to forged documents – leading to billions in fraud losses and compliance fines each year. Traditional static test datasets cannot keep up with these adversarial techniques. MIID addresses this gap by continuously generating “inorganic” identities (artificial identities with realistic attributes) that can be used to probe, test, and fortify anti-fraud and AML (Anti-Money Laundering) measures. In practical terms, MIID functions as the identity-centric testing backbone of the Yanez Compliance platform, ensuring that tools like sanctions screening and customer verification are resilient against identity manipulation tricks.
Bittensor Subnet 54 (MIID) is a specialized decentralized network designed by Yanez Compliance to generate synthetic “inorganic” identities for testing and strengthening financial crime prevention systems. In essence, MIID serves as a data-generation engine that creates realistic, varied identity profiles (names, documents, biometrics, etc.) to stress-test compliance tools (e.g. sanctions screening, KYC verification, fraud detection) under adversarial conditions. Unlike many experimental AI subnets, MIID is backed by an established compliance technology company with real enterprise use-cases, ensuring the project is grounded in practical demand and not just speculation. Below is a comprehensive breakdown of MIID’s purpose, operation, architecture, team, and roadmap.
MIID’s core purpose is to produce high-quality synthetic identity data to improve and validate financial crime defense systems. Financial institutions face ever-evolving evasion tactics – from fake identities to forged documents – leading to billions in fraud losses and compliance fines each year. Traditional static test datasets cannot keep up with these adversarial techniques. MIID addresses this gap by continuously generating “inorganic” identities (artificial identities with realistic attributes) that can be used to probe, test, and fortify anti-fraud and AML (Anti-Money Laundering) measures. In practical terms, MIID functions as the identity-centric testing backbone of the Yanez Compliance platform, ensuring that tools like sanctions screening and customer verification are resilient against identity manipulation tricks.
Key functional goals of Subnet 54 – MIID include:
Simulating Risk Scenarios: Create name variations, transliterations, and other identity mutations to mimic how bad actors evade detection (e.g. slight spelling changes to avoid sanctions list matches). By testing systems with these variations, MIID helps uncover weaknesses in name-matching and screening algorithms.
Training and Evaluation Data: Provide a live stream of diverse synthetic identities (fake personas, documents, biometrics) for model training and validation. This enables banks and researchers to improve their fraud detection and KYC models without using real customer data (preserving privacy and compliance).
Comprehensive Identity Testing: Evolve into a multimodal dataset covering names, documents, biometrics, and behavior. The subnet initially focuses on name-based adversarial cases (e.g. variant spellings, cultural naming differences), but it is designed to expand into generating synthetic fingerprints, faces, voices for anti-spoofing tests, as well as forged IDs and cross-modal identity scenarios. Ultimately, MIID will serve as a holistic identity intelligence source to test how well systems detect fraud across all aspects of an identity.
By fulfilling these functions, MIID aims to make financial compliance systems more adaptive and robust. Every synthetic identity produced is essentially a test case to ensure real criminal tactics are identified in simulation rather than in the real world. In short, MIID turns the creativity of fraudsters into a defense tool – it generates the kinds of fake identities criminals use, but for the benevolent purpose of hardening financial institutions’ defenses.
How It Works
MIID operates on Bittensor’s decentralized AI network, leveraging a miner–validator paradigm to produce and curate identity data in real time. The subnet (network UID 54 in Bittensor) consists of two main participant roles:
Miners (Data Producers): MIID miners run AI models that generate identity variations and synthetic profiles on demand. They receive challenge tasks (queries) from validators – for example, “produce variations of this name” or “simulate an ID document with specific traits.” Each miner uses large language models (LLMs) and other algorithms to create valid yet challenging identity data that fits the task. For instance, a miner might apply phonetic changes to “Alexander Ivanov” to produce variants like “Aleksandr Ivanoff,” or transliterate a name into another script, or fabricate a plausible fake passport with subtle errors. Miners then submit their generated identities back to the network. Their goal is to be as accurate, realistic, and diverse as possible, since better outputs earn higher rewards.
Validators (Data Curators): MIID validators act as quality control and judges for the miners’ submissions. They continuously issue new challenge prompts to miners and then evaluate the returned identity data on several criteria: linguistic accuracy (does the variant follow real-world language rules?), realism (would this fake identity plausibly evade detection?), and relevance to the intended test scenario. Validators score each miner’s response and rank miners based on the utility of their outputs. High-quality contributions (e.g. cleverly modified names that would trick naive systems, or well-crafted fake documents) are rewarded, while low-value or repetitive data can be penalized. Validators thus ensure the dataset remains robust and realistic, continuously curating the best synthetic identities for inclusion in the MIID dataset.
This miner–validator loop is incentivized by Bittensor’s crypto-economic system. Bittensor (the underlying protocol) rewards useful AI work with TAO tokens. In MIID’s case, miners whose identity outputs are frequently selected (i.e. provide real testing utility) receive more TAO, and validators earn rewards for diligently vetting data. This creates a competitive dynamic: many independent AI models “compete” to imagine new identity evasions, and only the most innovative or accurate ones get rewarded. As a result, MIID yields a continuous stream of adversarial identity data that evolves as fast as emerging threats. The Yanez Compliance platform (the company’s enterprise product) directly consumes this stream by integrating the vetted synthetic identities into its testing workflows. This means, for example, a bank using Yanez’s software can automatically test its sanction screening system each day with the latest fake names generated by MIID, or test its onboarding KYC with the newest AI-generated fake IDs.
Notably, MIID’s design creates a virtuous feedback loop between business and network: As Yanez Compliance’s clients use the data to validate their systems, the insights (which identity tricks slipped through, which were caught) inform the subnet’s future queries. Moreover, Yanez has committed to feed a portion of its commercial revenue back into the subnet – miners and validators are thus indirectly rewarded based on real-world usage of the data they create. This alignment ensures that what miners produce isn’t just theoretically interesting, but directly valuable to financial institutions’ needs. In summary, MIID works by harnessing decentralized AI to generate ever-improving fake identities, rigorously filtering them through validation, and then plugging them into real compliance tests – all under an incentive model that ties rewards to real utility.
Technical Architecture
MIID is built as a dedicated Bittensor subnet (network ID 54) with a focus on identity-simulation tasks. Architecturally, it inherits the Bittensor protocol’s core components and adds custom logic for the compliance domain:
Bittensor Protocol Foundation: The subnet runs on Bittensor’s decentralized network infrastructure, meaning it uses the standard miner/validator framework and TAO token incentives of Bittensor. All participating nodes run the open-source MIID software (available in the yanez-compliance/MIID-subnet repository) which extends Bittensor’s base classes for mining and validating. Communication between miners and validators occurs over the Bittensor peer-to-peer network (port 8091 is used for subnet 54’s miner-validator communication). Because it is a subnet, MIID operates within a specified domain of expertise (identity data) and has its own portion of the Bittensor metagraph where it can issue subtensor blocks and rewards specific to netuid 54.
Identity Generation Pipeline: On the miner side, the software pipeline involves LLMs and rule-based generators for identity data. According to the project docs, MIID miners use large language models (the default setup uses an Ollama instance with a LLaMA 3.1 model by default) to generate name variations and other identity attributes. These models are guided by predefined transformation rules and threat scenario specifications (for example, phonetic alterations, adding/removing middle names, altering dates or locations). The architecture allows inclusion of various data modalities: for future phases, miners will incorporate computer vision or speech models to generate synthetic biometrics and documents as well. All generated data is structured and returned in response to validator challenges.
Real-time Validation and Data Storage: Validators in subnet 54 maintain the logic for evaluating miner outputs. Architecturally, validators likely embed language-processing components (to check linguistic accuracy of names) and possibly verification heuristics (to simulate how a compliance system would react to the fake identity). They then use Bittensor’s consensus mechanisms to score and prune the data. High-scoring synthetic identities get added to the subnet’s growing dataset (which can be thought of as a continually updated corpus of adversarial identities). This dataset is accessible for training and testing AI models. Each block interval, rewards are distributed based on these scores, using Bittensor’s native on-chain incentive model.
Deployment and Integration: MIID was deployed on the Bittensor mainnet in May 2025 as per its roadmap. Initially it operates as a closed-loop test network (Yanez refers to it as a “test network” powering their platform), but by design it can integrate with external systems. Phase 4 of the roadmap (see below) includes opening the subnet to direct queries from institutions and regulators – implying the architecture will expose APIs or interfaces for outside entities to submit their own test scenarios to the network. Already, Yanez’s internal compliance platform taps into MIID’s output via custom connectors.
Economic and Governance Model: A distinguishing architectural feature of MIID is its hybrid incentive model. While it uses Bittensor’s token emission for miner/validator rewards, the team has explicitly addressed long-term sustainability by securing external funding and committing product revenues to support the subnet. This is atypical among Bittensor subnets – MIID is “built for sustainability, not speculation”, meaning it isn’t solely relying on token inflation or hype. Yanez raised capital (with support from partners like Yuma – an accelerator and investor in Bittensor projects backed by Digital Currency Group) to fund development and early operations. As the Yanez Compliance business grows, a portion of its revenue (from clients using the compliance platform) is funneled back to incentivize subnet participants. Technically, this could mean periodic injections of funding or buy-back of TAO to reward contributors. Governance-wise, Yanez maintains a degree of oversight (to ensure quality and compliance requirements are met), but in line with decentralized principles, the subnet’s evolution (e.g. which threat scenarios to prioritize, how scoring works) will increasingly be governed by transparent rules and possibly community input. The project emphasizes fiduciary responsibility and utility-driven development, indicating that every architectural decision is measured against real-world performance in detecting financial crime.
In summary, MIID’s architecture is a fusion of decentralized AI computing with compliance industry expertise. It uses Bittensor’s robust AI network framework, enhanced with specialized modules for identity simulation, and ties into a commercial platform. This allows MIID to operate at scale (decentralized miners provide potentially massive computational power), with adaptability (independent models contribute diverse approaches), and with a feedback loop to real-world outcomes (through Yanez’s platform and reinvested revenue). The result is an ever-evolving “identity adversary” system: a network that constantly learns how to fake identities in order to make the financial system safer.
Key functional goals of Subnet 54 – MIID include:
Simulating Risk Scenarios: Create name variations, transliterations, and other identity mutations to mimic how bad actors evade detection (e.g. slight spelling changes to avoid sanctions list matches). By testing systems with these variations, MIID helps uncover weaknesses in name-matching and screening algorithms.
Training and Evaluation Data: Provide a live stream of diverse synthetic identities (fake personas, documents, biometrics) for model training and validation. This enables banks and researchers to improve their fraud detection and KYC models without using real customer data (preserving privacy and compliance).
Comprehensive Identity Testing: Evolve into a multimodal dataset covering names, documents, biometrics, and behavior. The subnet initially focuses on name-based adversarial cases (e.g. variant spellings, cultural naming differences), but it is designed to expand into generating synthetic fingerprints, faces, voices for anti-spoofing tests, as well as forged IDs and cross-modal identity scenarios. Ultimately, MIID will serve as a holistic identity intelligence source to test how well systems detect fraud across all aspects of an identity.
By fulfilling these functions, MIID aims to make financial compliance systems more adaptive and robust. Every synthetic identity produced is essentially a test case to ensure real criminal tactics are identified in simulation rather than in the real world. In short, MIID turns the creativity of fraudsters into a defense tool – it generates the kinds of fake identities criminals use, but for the benevolent purpose of hardening financial institutions’ defenses.
How It Works
MIID operates on Bittensor’s decentralized AI network, leveraging a miner–validator paradigm to produce and curate identity data in real time. The subnet (network UID 54 in Bittensor) consists of two main participant roles:
Miners (Data Producers): MIID miners run AI models that generate identity variations and synthetic profiles on demand. They receive challenge tasks (queries) from validators – for example, “produce variations of this name” or “simulate an ID document with specific traits.” Each miner uses large language models (LLMs) and other algorithms to create valid yet challenging identity data that fits the task. For instance, a miner might apply phonetic changes to “Alexander Ivanov” to produce variants like “Aleksandr Ivanoff,” or transliterate a name into another script, or fabricate a plausible fake passport with subtle errors. Miners then submit their generated identities back to the network. Their goal is to be as accurate, realistic, and diverse as possible, since better outputs earn higher rewards.
Validators (Data Curators): MIID validators act as quality control and judges for the miners’ submissions. They continuously issue new challenge prompts to miners and then evaluate the returned identity data on several criteria: linguistic accuracy (does the variant follow real-world language rules?), realism (would this fake identity plausibly evade detection?), and relevance to the intended test scenario. Validators score each miner’s response and rank miners based on the utility of their outputs. High-quality contributions (e.g. cleverly modified names that would trick naive systems, or well-crafted fake documents) are rewarded, while low-value or repetitive data can be penalized. Validators thus ensure the dataset remains robust and realistic, continuously curating the best synthetic identities for inclusion in the MIID dataset.
This miner–validator loop is incentivized by Bittensor’s crypto-economic system. Bittensor (the underlying protocol) rewards useful AI work with TAO tokens. In MIID’s case, miners whose identity outputs are frequently selected (i.e. provide real testing utility) receive more TAO, and validators earn rewards for diligently vetting data. This creates a competitive dynamic: many independent AI models “compete” to imagine new identity evasions, and only the most innovative or accurate ones get rewarded. As a result, MIID yields a continuous stream of adversarial identity data that evolves as fast as emerging threats. The Yanez Compliance platform (the company’s enterprise product) directly consumes this stream by integrating the vetted synthetic identities into its testing workflows. This means, for example, a bank using Yanez’s software can automatically test its sanction screening system each day with the latest fake names generated by MIID, or test its onboarding KYC with the newest AI-generated fake IDs.
Notably, MIID’s design creates a virtuous feedback loop between business and network: As Yanez Compliance’s clients use the data to validate their systems, the insights (which identity tricks slipped through, which were caught) inform the subnet’s future queries. Moreover, Yanez has committed to feed a portion of its commercial revenue back into the subnet – miners and validators are thus indirectly rewarded based on real-world usage of the data they create. This alignment ensures that what miners produce isn’t just theoretically interesting, but directly valuable to financial institutions’ needs. In summary, MIID works by harnessing decentralized AI to generate ever-improving fake identities, rigorously filtering them through validation, and then plugging them into real compliance tests – all under an incentive model that ties rewards to real utility.
Technical Architecture
MIID is built as a dedicated Bittensor subnet (network ID 54) with a focus on identity-simulation tasks. Architecturally, it inherits the Bittensor protocol’s core components and adds custom logic for the compliance domain:
Bittensor Protocol Foundation: The subnet runs on Bittensor’s decentralized network infrastructure, meaning it uses the standard miner/validator framework and TAO token incentives of Bittensor. All participating nodes run the open-source MIID software (available in the yanez-compliance/MIID-subnet repository) which extends Bittensor’s base classes for mining and validating. Communication between miners and validators occurs over the Bittensor peer-to-peer network (port 8091 is used for subnet 54’s miner-validator communication). Because it is a subnet, MIID operates within a specified domain of expertise (identity data) and has its own portion of the Bittensor metagraph where it can issue subtensor blocks and rewards specific to netuid 54.
Identity Generation Pipeline: On the miner side, the software pipeline involves LLMs and rule-based generators for identity data. According to the project docs, MIID miners use large language models (the default setup uses an Ollama instance with a LLaMA 3.1 model by default) to generate name variations and other identity attributes. These models are guided by predefined transformation rules and threat scenario specifications (for example, phonetic alterations, adding/removing middle names, altering dates or locations). The architecture allows inclusion of various data modalities: for future phases, miners will incorporate computer vision or speech models to generate synthetic biometrics and documents as well. All generated data is structured and returned in response to validator challenges.
Real-time Validation and Data Storage: Validators in subnet 54 maintain the logic for evaluating miner outputs. Architecturally, validators likely embed language-processing components (to check linguistic accuracy of names) and possibly verification heuristics (to simulate how a compliance system would react to the fake identity). They then use Bittensor’s consensus mechanisms to score and prune the data. High-scoring synthetic identities get added to the subnet’s growing dataset (which can be thought of as a continually updated corpus of adversarial identities). This dataset is accessible for training and testing AI models. Each block interval, rewards are distributed based on these scores, using Bittensor’s native on-chain incentive model.
Deployment and Integration: MIID was deployed on the Bittensor mainnet in May 2025 as per its roadmap. Initially it operates as a closed-loop test network (Yanez refers to it as a “test network” powering their platform), but by design it can integrate with external systems. Phase 4 of the roadmap (see below) includes opening the subnet to direct queries from institutions and regulators – implying the architecture will expose APIs or interfaces for outside entities to submit their own test scenarios to the network. Already, Yanez’s internal compliance platform taps into MIID’s output via custom connectors.
Economic and Governance Model: A distinguishing architectural feature of MIID is its hybrid incentive model. While it uses Bittensor’s token emission for miner/validator rewards, the team has explicitly addressed long-term sustainability by securing external funding and committing product revenues to support the subnet. This is atypical among Bittensor subnets – MIID is “built for sustainability, not speculation”, meaning it isn’t solely relying on token inflation or hype. Yanez raised capital (with support from partners like Yuma – an accelerator and investor in Bittensor projects backed by Digital Currency Group) to fund development and early operations. As the Yanez Compliance business grows, a portion of its revenue (from clients using the compliance platform) is funneled back to incentivize subnet participants. Technically, this could mean periodic injections of funding or buy-back of TAO to reward contributors. Governance-wise, Yanez maintains a degree of oversight (to ensure quality and compliance requirements are met), but in line with decentralized principles, the subnet’s evolution (e.g. which threat scenarios to prioritize, how scoring works) will increasingly be governed by transparent rules and possibly community input. The project emphasizes fiduciary responsibility and utility-driven development, indicating that every architectural decision is measured against real-world performance in detecting financial crime.
In summary, MIID’s architecture is a fusion of decentralized AI computing with compliance industry expertise. It uses Bittensor’s robust AI network framework, enhanced with specialized modules for identity simulation, and ties into a commercial platform. This allows MIID to operate at scale (decentralized miners provide potentially massive computational power), with adaptability (independent models contribute diverse approaches), and with a feedback loop to real-world outcomes (through Yanez’s platform and reinvested revenue). The result is an ever-evolving “identity adversary” system: a network that constantly learns how to fake identities in order to make the financial system safer.
MIID (Subnet 54) is developed and maintained by the team at Yanez Compliance, Inc., with support from partners in the Bittensor ecosystem. Key publicly known contributors and affiliations include:
Jose Caldera – CEO: Jose is the chief executive leading the vision for Yanez and its MIID subnet. He has spoken about MIID as a transformative approach to financial crime prevention, drawing parallels to continuous vulnerability scanning in cybersecurity. Under his leadership, Yanez has positioned MIID as core infrastructure to help banks and institutions prove their compliance systems work against daily-evolving threats.
Bin Tang – CTO: Strong development experience in enterprise software, web application, and project management. Experienced in Java, Ruby, SQL, Javascript, Cassandra, Oracle, MySQL, Hibernate, Spring, Rails, Mongo DB, Data Analysis, Machine Learning, AWS, Apache Spark, Solr, Graph DB, Python, R.
Asem Othman – Chief AI Officer (CAIO): A seasoned AI and biometric identity expert who leads the technical development of the MIID subnet. He is a top contributor to the project’s open-source codebase, and brings 15+ years of experience in machine learning applied to identity verification. Under his guidance, the subnet’s AI models and threat scenario logic have been developed. (Affiliation: CAIO at Yanez Compliance – Palo Alto).
Mark Quesenberry – Head of Sales: Rich experience in net new market creation and building successful sales/business development teams, including organizational strategy, execution, and business process improvement.
Danielle Z – UX/UI Designer: Experience creating low and high-fidelity wireframes, prototypes, graphics, and visual design elements. Proficient with design tools like Figma, Adobe XD, and Illustrator.
Yossi Zekri – Strategic Advisor & Investor
David Anwyl – Research Intern
MIID (Subnet 54) is developed and maintained by the team at Yanez Compliance, Inc., with support from partners in the Bittensor ecosystem. Key publicly known contributors and affiliations include:
Jose Caldera – CEO: Jose is the chief executive leading the vision for Yanez and its MIID subnet. He has spoken about MIID as a transformative approach to financial crime prevention, drawing parallels to continuous vulnerability scanning in cybersecurity. Under his leadership, Yanez has positioned MIID as core infrastructure to help banks and institutions prove their compliance systems work against daily-evolving threats.
Bin Tang – CTO: Strong development experience in enterprise software, web application, and project management. Experienced in Java, Ruby, SQL, Javascript, Cassandra, Oracle, MySQL, Hibernate, Spring, Rails, Mongo DB, Data Analysis, Machine Learning, AWS, Apache Spark, Solr, Graph DB, Python, R.
Asem Othman – Chief AI Officer (CAIO): A seasoned AI and biometric identity expert who leads the technical development of the MIID subnet. He is a top contributor to the project’s open-source codebase, and brings 15+ years of experience in machine learning applied to identity verification. Under his guidance, the subnet’s AI models and threat scenario logic have been developed. (Affiliation: CAIO at Yanez Compliance – Palo Alto).
Mark Quesenberry – Head of Sales: Rich experience in net new market creation and building successful sales/business development teams, including organizational strategy, execution, and business process improvement.
Danielle Z – UX/UI Designer: Experience creating low and high-fidelity wireframes, prototypes, graphics, and visual design elements. Proficient with design tools like Figma, Adobe XD, and Illustrator.
Yossi Zekri – Strategic Advisor & Investor
David Anwyl – Research Intern
MIID’s development roadmap is outlined in distinct phases, spanning from its recent launch in 2025 through expansions into 2026–2027. Each phase adds new capabilities and broadens the subnet’s scope in generating and testing identity data. Below are the planned phases, milestones, and timelines (as currently published):
Phase 1: Initial Launch & Name-Based Threat Scenarios (May 2025) – Launch of Subnet 54 on Bittensor mainnet, focusing on name variations as the first attack vector. In this phase, validators use known “threat scenarios” (e.g. known tricky name variants) to challenge miners. The subnet generates phonetic and orthographic name variations (e.g. missing diacritics, look-alike characters) and other rule-based permutations to test sanctions screening systems. This established the foundation and proved the concept with name-centric data.
Phase 2: Miner-Contributed Threat Scenarios (June 2025) – Introduction of a more dynamic querying system. Miners are now allowed to propose new or unknown threat scenarios themselves, rather than only validators posing pre-set challenges. A “post-evaluation system” is implemented so that when miners suggest novel identity manipulations, validators systematically validate and incorporate them if useful. This phase also broadens the type of evasion tactics: beyond phonetic/orthographic changes, it includes nickname-based variations, transliteration alterations, and middle-name manipulations as additional ways criminals might obscure identities. Validator scoring is refined to handle this greater variety, and penalties are introduced for miners who spam low-value or duplicate data (encouraging creativity over quantity).
Phase 3: Location-Based Threat Scenarios (August 2025) – Expansion of the subnet’s capabilities to handle geographic and location data in identities. In this stage, MIID will generate and test scenarios where bad actors hide or fake location-related identity attributes. For example, adding support for location obfuscation: miners might produce identities with inconsistent or misleading address information, or simulate users masking their IP or residence in sanctioned regions. Validators will include queries tied to high-risk regions (e.g. “flag if identity appears to be from X country”) to see if systems can catch location-based evasion. This enhances testing for sanctions and AML systems that must track geographic risk.
Phase 4: AML Ecosystem Integration (September 2025) – Opening the subnet to external stakeholders and real-world queries. By this point, the MIID subnet will not just serve Yanez’s internal platform, but also allow financial institutions and regulators to interface directly. External institutions could submit their own compliance questions or test cases into the subnet and get results powered by the miners. The subnet’s “execution vectors” (types of identity tests) are expanded and validated to meet enterprise-grade standards. Essentially, MIID transitions from a Yanez-operated testbed into a more open service for the broader compliance community, marking a key milestone of industry adoption.
Phase 5–11: Identity Realism & Simulation (2026–2027) – A series of enhancements aimed at multi-modal identity simulation to achieve high realism. Planned milestones during 2026 include integrating biometric data generation by Q1 2026 (e.g. synthetic fingerprints, facial images), and AI-generated documents by Q2 2026 (e.g. fake passports or ID cards with controlled anomalies). By Q3 2026, the subnet will simulate digital presence and interactions – potentially creating fictitious social media or transaction behavior tied to identities, adding a behavioral dimension to testing. Q4 2026 introduces financial transaction modeling, so the subnet can generate fake transaction patterns linked to synthetic identities (useful for testing transaction monitoring systems). Moving into 2027, by Q2 2027 the plan is to build 3D identity “avatars” – perhaps virtual persons with profiles across multiple platforms – and to add voice and conversational AI support (no date given, but listed in this phase). These steps will push MIID to truly multimodal AI generation, covering text, image, biometric, and voice data for identities.
Final Phase: Unified Identity Representation (post-2027) – The ultimate vision for MIID is to culminate in a unified AI model for identity screening and a collaborative validation platform. In this final phase, the project aims to train a comprehensive model that encapsulates all the knowledge gained – essentially an AI that can assess identity risk holistically. Additionally, a decentralized platform will be launched for ongoing collaborative validation and contribution, indicating a fully open system where participants (possibly including institutions, researchers, and decentralized AI nodes) continuously improve the identity screening model together. This suggests MIID could evolve from just a subnet into a broader ecosystem or service for identity risk intelligence.
Throughout these phases, MIID’s roadmap shows a clear trajectory: from generating simple name tweaks to simulating entire synthetic personas interacting in a financial system. Each milestone builds on the last, steadily approaching the goal of a comprehensive “adversarial identity” simulator for the financial industry. The team has also indicated ongoing “future plans” such as incorporating more complex attributes (addresses, DOBs) and improving multi-modal AI performance – underscoring that the roadmap is dynamic and will adapt as new challenges in fraud prevention emerge.
Current Status: As of mid-2025, MIID Subnet 54 has launched and is in its early phases of operation. The test network is live with miners and validators contributing name variation data, and Yanez is actively inviting participants (via Discord and other channels) to mine, validate, and collaborate. The initial focus on sanctions screening validation is already in use with Yanez’s clients, demonstrating real demand. Future phases are expected to roll out according to the roadmap, though specific dates might adjust as development progresses.
MIID’s development roadmap is outlined in distinct phases, spanning from its recent launch in 2025 through expansions into 2026–2027. Each phase adds new capabilities and broadens the subnet’s scope in generating and testing identity data. Below are the planned phases, milestones, and timelines (as currently published):
Phase 1: Initial Launch & Name-Based Threat Scenarios (May 2025) – Launch of Subnet 54 on Bittensor mainnet, focusing on name variations as the first attack vector. In this phase, validators use known “threat scenarios” (e.g. known tricky name variants) to challenge miners. The subnet generates phonetic and orthographic name variations (e.g. missing diacritics, look-alike characters) and other rule-based permutations to test sanctions screening systems. This established the foundation and proved the concept with name-centric data.
Phase 2: Miner-Contributed Threat Scenarios (June 2025) – Introduction of a more dynamic querying system. Miners are now allowed to propose new or unknown threat scenarios themselves, rather than only validators posing pre-set challenges. A “post-evaluation system” is implemented so that when miners suggest novel identity manipulations, validators systematically validate and incorporate them if useful. This phase also broadens the type of evasion tactics: beyond phonetic/orthographic changes, it includes nickname-based variations, transliteration alterations, and middle-name manipulations as additional ways criminals might obscure identities. Validator scoring is refined to handle this greater variety, and penalties are introduced for miners who spam low-value or duplicate data (encouraging creativity over quantity).
Phase 3: Location-Based Threat Scenarios (August 2025) – Expansion of the subnet’s capabilities to handle geographic and location data in identities. In this stage, MIID will generate and test scenarios where bad actors hide or fake location-related identity attributes. For example, adding support for location obfuscation: miners might produce identities with inconsistent or misleading address information, or simulate users masking their IP or residence in sanctioned regions. Validators will include queries tied to high-risk regions (e.g. “flag if identity appears to be from X country”) to see if systems can catch location-based evasion. This enhances testing for sanctions and AML systems that must track geographic risk.
Phase 4: AML Ecosystem Integration (September 2025) – Opening the subnet to external stakeholders and real-world queries. By this point, the MIID subnet will not just serve Yanez’s internal platform, but also allow financial institutions and regulators to interface directly. External institutions could submit their own compliance questions or test cases into the subnet and get results powered by the miners. The subnet’s “execution vectors” (types of identity tests) are expanded and validated to meet enterprise-grade standards. Essentially, MIID transitions from a Yanez-operated testbed into a more open service for the broader compliance community, marking a key milestone of industry adoption.
Phase 5–11: Identity Realism & Simulation (2026–2027) – A series of enhancements aimed at multi-modal identity simulation to achieve high realism. Planned milestones during 2026 include integrating biometric data generation by Q1 2026 (e.g. synthetic fingerprints, facial images), and AI-generated documents by Q2 2026 (e.g. fake passports or ID cards with controlled anomalies). By Q3 2026, the subnet will simulate digital presence and interactions – potentially creating fictitious social media or transaction behavior tied to identities, adding a behavioral dimension to testing. Q4 2026 introduces financial transaction modeling, so the subnet can generate fake transaction patterns linked to synthetic identities (useful for testing transaction monitoring systems). Moving into 2027, by Q2 2027 the plan is to build 3D identity “avatars” – perhaps virtual persons with profiles across multiple platforms – and to add voice and conversational AI support (no date given, but listed in this phase). These steps will push MIID to truly multimodal AI generation, covering text, image, biometric, and voice data for identities.
Final Phase: Unified Identity Representation (post-2027) – The ultimate vision for MIID is to culminate in a unified AI model for identity screening and a collaborative validation platform. In this final phase, the project aims to train a comprehensive model that encapsulates all the knowledge gained – essentially an AI that can assess identity risk holistically. Additionally, a decentralized platform will be launched for ongoing collaborative validation and contribution, indicating a fully open system where participants (possibly including institutions, researchers, and decentralized AI nodes) continuously improve the identity screening model together. This suggests MIID could evolve from just a subnet into a broader ecosystem or service for identity risk intelligence.
Throughout these phases, MIID’s roadmap shows a clear trajectory: from generating simple name tweaks to simulating entire synthetic personas interacting in a financial system. Each milestone builds on the last, steadily approaching the goal of a comprehensive “adversarial identity” simulator for the financial industry. The team has also indicated ongoing “future plans” such as incorporating more complex attributes (addresses, DOBs) and improving multi-modal AI performance – underscoring that the roadmap is dynamic and will adapt as new challenges in fraud prevention emerge.
Current Status: As of mid-2025, MIID Subnet 54 has launched and is in its early phases of operation. The test network is live with miners and validators contributing name variation data, and Yanez is actively inviting participants (via Discord and other channels) to mine, validate, and collaborate. The initial focus on sanctions screening validation is already in use with Yanez’s clients, demonstrating real demand. Future phases are expected to roll out according to the roadmap, though specific dates might adjust as development progresses.
Strong start to the month. Committed. Are you paying attention?
#SN54 #Yanez #Bittensor
@micaelabazo speaking of sn collaboration as no1 listened to their spaces. @yanez__ai #SN54 working with multiple sn's and in talks with way more, and getting feedback like this; https://x.com/yubrew/status/1940137071328329920
Testing sanctions screening models on clean, static data won’t reveal its blind spots.
Yanez MIID feeds systems anomalies, edge cases, and real-world tactics—generated on demand by competing miners of @bittensor_.
#SN54
Investor community in NYC. We'll be in NYC from July 14-16. DM us if you are interested in having a sit down conversation about SN54, our business, investing in subnets and #bittensor in general.
Bittensor Ecosystem Updates
Mainnet launches and major milestones
@yanez__ai (SN60) is partnering with @bitsecai to strengthen their incentive algorithm. Cross-subnet collaboration is becoming the standard.
@metanova_labs (SN68) achieved 4.6 million molecule submissions against…
Come and join us for coffee and pizza, and one of the best views in San Francisco. @btlabs_ai and us will be sharing our experiences in setting up a subnet, how to integrate Bittensor into a W2 company, and much more. See you there. Everybody welcome.
Brunchtensor · Luma
Kickstart your day with BTLabs, Yanez.ai, and a vibrant crew of Bittensor builders, investors, and enthusiasts. Come ...
lu.ma
Keep ahead of the Bittensor exponential development curve…
Subnet Alpha is an informational platform for Bittensor Subnets.
This site is not affiliated with the Opentensor Foundation or TaoStats.
The content provided on this website is for informational purposes only. We make no guarantees regarding the accuracy or currency of the information at any given time.
Subnet Alpha is created and maintained by The Realistic Trader. If you have any suggestions or encounter any issues, please contact us at [email protected].
Copyright 2024
