The Intersection of AI and Crypto

The artificial intelligence industry has been making headlines lately, for reasons good and bad. While you’re probably well aware of the recent drama surrounding OpenAI and have maybe explored the capabilities of existing artificial intelligence technology, you probably haven’t thought much about how artificial intelligence can interact with blockchain-based systems. In this week’s report we’ll be covering a handful of existing applications attempting to leverage both artificial intelligence and blockchain technology, along with some information concerning the future of these apps and the artificial intelligence industry in the years to come.


Key Takeaways

  • AI has recently been making waves in the tech industry, but many may not yet be familiar with the role crypto can play in enabling AI applications and help solve some of the pertinent challenges faced by AI today. 
  • One of the key areas is in data management and security. AI systems require vast amounts of data to learn and improve. Through blockchains, training data can be shared across different platforms and stakeholders, enabling more avenues for collaboration in AI research and development. Crypto can also be used to incentivize the sharing of AI data to foster a more inclusive ecosystem.

  • AI can in the future power DAOs, which currently still require human participation and financial incentives. AI can lead to the creation of truly fully automated DAOs.

  • Crypto can also lead to the creation of decentralized AI models and democratize access to AI technologies, allowing individuals and small entities to access AI tools and services that were previously the domain of large corporations.

  • The report highlights several projects, such as Bittensor, Akash Network, Render, Gensyn Network and as prominent examples of projects working at the intersection of AI and crypto.


What is AI and how does it relate to crypto?

Before we get into project specifics and some of the more technical details, it’s important to cover some of the basics around artificial intelligence technology and how the talented teams and individual developers within the industry have got us to today.

There’s a strong chance you are already familiar with ChatGPT, the most popular and widely recognized consumer-facing AI application that has consumed the tech industry’s attention over the last year – today we’ll briefly explain the concepts underpinning this technology and how it’s able to perform so competently at, well, basically everything it’s asked to compute.

The core piece of technology powering ChatGPT and other consumer chat-based models is what’s known as the large language model, otherwise known as an LLM. These complicated pieces of AI tech are essentially a combination of deep learning techniques / algorithms and very large data sets that work together to create an artificial intelligence model capable of predicting and summarizing knowledge.

Interactions between humans and LLMs are handled via natural language, with most LLMs being built specifically with natural language processing (NLP) in mind. A user asks a chatbot to answer some type of question in natural language, with the chatbot then using its underlying technology, training data and capabilities to provide an answer to the user as best as it can.

LLMs are built upon transformer models, commonly referred to as transformers. These are a type of neural network that excel at predicting text and learning the context behind words. Because LLMs that use transformer models excel at NLP, they’re able to work really well for common tasks that humans need everyday, things like solving math problems, generating code templates or even writing shorter reports or suggesting edits.

Because of this, chatbots like ChatGPT, Microsoft Bing AI and Claude have seen immense success and have almost single handedly sparked an AI revolution. While many believe that AI systems might eventually gain capabilities and intelligence greater than humans, there is little evidence to suggest that this will happen anytime soon. Regardless, the possibilities that come from these models integrating with human workflows and the extremely promising existing capabilities prove that AI is here to stay, whether we all like it or not. But you’re probably wondering how these models can fit in with crypto and the permissionless nature of blockchains, so let’s explain the potential synchronicities and examine these two radical forms of technology.


How can crypto help enable AI applications?

The crypto industry is one that’s consistently discussed on the news, in large media outlets and across other social media platforms every single day. What started with a single whitepaper written by Satoshi in 2008 has transformed into a $1.5 trillion market with a flurry of looming ETF approvals or denials from the largest financial institutions in the world.

It’s often difficult to describe the innate benefits of blockchain technology to industry outsiders, mainly because the financial industry is very well developed and smooth in the majority of first world countries. Outside of places like the United States, it’s much easier to explain and show the power of permissionless ledgers for financial transactions, largely due to the corrupt financial institutions and governments that unfortunately still hold power in every area of the world. Currencies are regularly getting debased in nations across the globe, with a large majority of the world’s population still without access to banking infrastructure that’s often seen as an afterthought in the United States.

Crypto is a way of banking the unbanked, a technology that offers an opportunity for individuals to become their own overseer of financial operations, whether they’re holding crypto in a cold storage wallet or utilizing the numerous decentralized finance applications available across the crypto ecosystem. The promise of permissionless finance can’t easily be described, but the revolution occurring each day cannot be understated.

A blockchain’s inherent characteristics of transparency, security, and decentralization can significantly contribute to the way AI data is stored, shared, and utilized. This amalgamation of technologies promises to enhance trust in AI systems by providing an immutable ledger for AI transactions and decisions, thereby reducing concerns over data manipulation or misuse.

One of the critical aspects where crypto can assist AI (and vice versa) is in the realm of data management and security. AI systems require vast amounts of data to learn and improve. By leveraging blockchain technology, this data can be securely and transparently shared across different platforms and stakeholders. This not only ensures the integrity of the data but also opens up new avenues for collaborative AI research and development, breaking down data silos that often hinder innovation.

The integration of AI and blockchain could lead to the creation of legitimately decentralized autonomous organizations (DAOs). These DAOs, governed by smart contracts and powered by AI algorithms, could operate independently, make decisions, and execute transactions without human intervention. Historically, DAO management incrypto has been less than ideal as human emotions and financial incentives often obscure the initial purposes of a DAO. Implementing AI systems could revolutionize industries by automating processes and reducing the need for intermediaries, thereby increasing efficiency and reducing costs.

Another promising area is the use of blockchains as a means to incentivize the generation and sharing of AI data. Through tokenization, individuals and organizations could be rewarded for contributing valuable data to AI models, fostering a more collaborative and inclusive AI ecosystem.

Decentralized finance (DeFi) is also a potentially huge benefactor of AI, potentially creating what could be referenced as decentralized AI (DeAI). This approach could democratize access to AI technologies, allowing individuals and small entities to access AI tools and services that were previously the domain of large corporations.

The convergence of cryptocurrency and AI holds the potential to transform not only the financial sector but many aspects of our digital lives. By combining the strengths of both technologies, we can look forward to a future where AI is not only more accessible, but more secure, transparent and potentially even more efficient. Speaking of, let’s breakdown the current workings of the AI industry and how it currently functions.


Breaking down the opaque walls of artificial intelligence

Comparing the overhaul of the financial system via crypto to the intelligence revolution occurring through the production of artificial intelligence systems, we can draw some very relevant similarities and make a case for the combination of the two.

In the present day, artificial intelligence companies like OpenAI, Google’s Deepmind, Anthropic and many, many others conduct their research and operations under closed doors.


Current opportunities in the crypto & artificial intelligence landscape

Now that we have covered some of the basics around AI and Crypto synergies, we can take a closer examination of some of the leading projects within the sector. While most of these are still actively working to bootstrap their networks, acquire a loyal user base and gain attention from the broader crypto community, they are all working at the forefront of the industry and represent a good representation of this rapidly growing sector.


Bittensor, a network of decentralized artificial intelligence models:

Bittensor is by far the most popular and well-established project building within the crypto & AI ecosystem. Bittensor is a decentralized network designed to democratize the field of artificial intelligence (AI) by creating a platform for numerous decentralized commodity markets, or ‘subnetworks’, united under a single token system. Its mission is to build a network that rivals the capabilities of large super corporations in AI, such as OpenAI, by employing unique incentive mechanisms and an advanced subnetwork architecture. Bittensor’s system can be thought of as a machine, facilitated by blockchains, to transfer AI capabilities on-chain efficiently.

The network is managed by two key players: miners and validators. Miners submit pre-trained AI models to the network and receive rewards for their contributions, while validators ensure the validity and accuracy of the models’ outputs. This setup creates a competitive environment where miners are incentivized to continually improve their models for better performance and greater rewards in $TAO, the network’s native token. Users interact with the network by sending queries to validators, who then distribute these to miners. The validators rank the outputs from these miners and return the highest-ranked responses to the user.

Bittensor’s approach to model development is unique. Unlike many AI labs or research organizations, Bittensor does not train models due to the high costs and complexity involved. Instead, the network relies on decentralized training mechanisms. Validators are tasked with evaluating the models produced by miners using a specific dataset and scoring each model based on certain criteria, such as accuracy and loss functions. This decentralized evaluation ensures a continuous improvement in model performance.

The architecture of Bittensor includes the Yuma Consensus mechanism, a unique hybrid of both Proof of Work (PoW) and Proof of Stake (PoS), which distributes resources across the network’s subnetworks. Subnetworks are self-contained economic markets each focusing on different AI tasks, like text prediction or image generation, and can choose to opt in or out of the Yuma Consensus depending on their functionality.

Bittensor represents a significant step in the decentralization of AI, offering a platform where diverse AI models can be developed, evaluated, and improved in a decentralized manner. Its unique structure not only incentivizes the creation of high-quality AI models but also democratizes access to AI technology, potentially transforming how AI is developed and used in various sectors.


Akash, an open-source supercloud:

The Akash Network is an innovative, open-source Supercloud platform designed for buying and selling computing resources in a secure and efficient manner. It is built with the vision of providing users the power to deploy their own cloud infrastructure as well as to buy and sell unused cloud resources. This flexibility not only democratizes cloud resource utilization but also offers cost-effective solutions for users needing to scale their operations.

At the core of Akash’s system is a reverse auction mechanism, where users can submit bids for their computing needs and providers compete to offer services, often resulting in significantly lower prices compared to traditional cloud systems. This system is underpinned by reliable and well-established technologies like Kubernetes and Cosmos, ensuring a secure and dependable platform for hosting applications. Akash’s community-driven approach ensures that its users have a say in the network’s development and governance, making it a truly public and user-centric service.

Akash’s infrastructure is defined using a simple-to-use, YAML-based Stack Definition Language (SDL), which allows users to create complex deployments across multiple areas and providers. This feature, combined with Kubernetes, the leading container orchestration system, guarantees not only flexibility in deployment but also security and reliability in application hosting. Furthermore, Akash offers persistent storage solutions, ensuring data retention even after restarts, which is particularly beneficial for applications managing large datasets.

Overall, Akash Network stands out as a decentralized cloud platform, offering a unique solution to the monopolistic nature of current cloud service providers. Its model of utilizing underutilized resources across millions of data centers globally not only reduces costs but also enhances the speed and efficiency of cloud-native applications. With no need for proprietary language rewrites and no vendor lock-in, Akash presents a versatile and accessible platform for a wide range of cloud-based applications.


Render, a platform for expanding access to compute:

The Render Network is a blockchain-based platform designed to address the growing computational demands in media production, particularly in fields like augmented reality, virtual reality, and AI-enhanced media. It leverages unused GPU cycles to connect content creators needing computational power with providers who have available GPU resources. This decentralized approach, facilitated through blockchain technology, ensures secure and efficient processing of GPU-based tasks, including AI-driven content creation and optimization.

Render Network’s core offering is its integration with AI, which plays a crucial role in both content creation and process optimization. The network supports AI-related tasks, enabling artists to use AI tools for generating assets and enhancing digital artwork. This integration allows for the creation of ultra-high resolution 3D worlds and optimized rendering processes, like AI denoising. Additionally, Render Network’s use of AI extends to managing large-scale art collections and optimizing the rendering workflow, thus broadening the possibilities in creative processes.

The ecosystem of Render Network functions as a marketplace for GPU resources, serving various stakeholders such as artists, engineers, and node operators. It democratizes access to computational power, enabling both individual creators and larger studios to undertake complex rendering projects affordably. Transactions within this ecosystem are facilitated using the RNDR token, creating a vibrant economy centered around rendering services. As AI continues to reshape digital content creation, the Render Network is poised to become a key player in facilitating new forms of creative expression and technological innovation in the digital media landscape.


Gensyn, a decentralized compute platform:

Gensyn is an AI and cryptocurrency project focused on addressing the computational challenges and resource limitations inherent in state-of-the-art Artificial Intelligence (AI) systems. The project aims to overcome the barriers to AI advancement caused by the enormous resource requirements needed to build foundational models. Gensyn’s approach is to create a decentralized, blockchain-based protocol for efficiently leveraging global compute resources.

The background of Gensyn highlights the increasing computational complexity of AI systems, which is outpacing the available compute supply. For instance, training large models like OpenAI’s GPT-4 requires substantial resources, creating significant barriers for all parties involved. This dynamic has led to demands for a system that can efficiently use all available compute resources, addressing the limitations of current solutions, which are either too expensive or insufficient for large-scale AI work.

Gensyn aims to solve this problem by creating a decentralized protocol that connects and verifies off-chain deep learning work in a cost-efficient manner. This protocol faces several challenges, including work verification, marketplace dynamics, ex-ante work estimation, privacy concerns, and the need for effective parallelization of deep learning models. The protocol intends to build a trustless compute network with economic incentives for participation and a method to verify that computational work has been performed as promised.

The Gensyn Protocol is a layer-1 trustless protocol for deep learning computation that rewards participants for contributing their compute time and performing ML tasks. It uses a combination of techniques to verify the work completed, including probabilistic proof-of-learning, a graph-based pinpoint protocol, and a Truebit-style incentive game. The system involves various participants such as Submitters, Solvers, Verifiers, and Whistleblowers, each playing a specific role in the computational process.

In practice, the Gensyn Protocol involves several stages, from task submission to contract arbitration and settlement. It aims to create a transparent, low-cost marketplace for ML compute, enabling scalability and efficiency. The protocol also offers an opportunity for miners with powerful GPUs to repurpose their hardware for ML computation, potentially at a lower cost compared to mainstream providers. This approach not only addresses the computational challenges of AI but also aims to democratize access to AI resources.


Fetch, an open platform for the artificial intelligence economy: has been around longer than some of the previously mentioned projects, with a large variety of services offered on its website. At its core, Fetch is an innovative project at the intersection of artificial intelligence (AI) and cryptocurrency, aimed at revolutionizing the way economic activities and processes are carried out. The foundation of Fetch’s offerings is its AI agents, designed as modular building blocks that can be programmed to execute specific tasks. These agents are capable of autonomously connecting, searching, and transacting, thereby creating dynamic marketplaces and altering the traditional landscape of economic activity.

One of the key services offered by Fetch is the ability to make legacy products AI-ready. This is achieved by integrating their APIs with Agents, a process that is quick and does not necessitate altering the underlying business application. The AI agents can be combined with other agents in the network, opening up possibilities for new use cases and business models. Furthermore, these agents possess the capability to negotiate and transact on behalf of users, enabling them to earn from their deployment.

Additionally, these agents can provide inferences from machine learning models, allowing users to monetize their insights and enhance their machine learning models.

Fetch also introduces Agentverse, a no-code managed service that simplifies the deployment of AI agents. Just like legacy no-code platforms are gaining traction (Replit)and services like Github’s Copilot making writing code accessible to the masses, Fetch is working to further democratize web3 development in its own unique way.

Through Agentverse, users can launch their first agent effortlessly, which significantly lowers the barrier to entry for using advanced AI technologies. In terms of AI Engine and Agent Services, Fetch utilizes large language models (LLMs) to discover and directtask execution to the appropriate AI agents. This system not only monetizes AI apps and services but also serves as a comprehensive platform for agent services including building, listing, analytics, and hosting.

The platform enhances its utility with features such as Search & Discovery and Analytics. Agents can be registered in the Agentverse for active discoverability on’s platform, which employs a targeted LLM-based search. Analytical tools are available for improving the effectiveness of an agent’s semantic descriptors, thereby enhancing their discoverability. Moreover, incorporates an IoT Gateway for offline agents, enabling them to collect messages and process them in batches upon reconnection.

Finally, offers hosting services for managed agents, providing all the features of Agentverse besides hosting. The platform also introduces an open network for agent addressing and naming, leveraging’s Web3 network. This aspect signifies a novel approach to Web DNS addressing, integrating blockchain technology into the system. Overall, presents a versatile platform that merges AI and blockchain technology, offering tools for AI agent development, machine learning model monetization, and a groundbreaking approach to search and discoverability in the digital economy. This combination of AI agents and blockchain technology paves the way for automating and optimizing various processes in a decentralized and efficient manner.


Next steps and projections for both industries

The seamless integration of artificial intelligence and blockchain technology represents a pivotal advancement in both sectors. This combination is not just a mere fusion of two cutting-edge technologies, but a transformational synergy that redefines the boundaries of digital innovation and decentralization. The potential applications of this integration, as explored in various projects like, Bittensor, Akash Network, Render Network, and Gensyn, demonstrate the vast possibilities and significant benefits of combining AI’s computational power with blockchain’s secure and transparent framework.

As we look toward the future, it is evident that the convergence of AI and blockchain will play a crucial role in shaping various industries. From enhancing data security and integrity to creating new models of decentralized autonomous organizations, this amalgamation holds the promise of more efficient, transparent, and accessible technologies. Particularly in the realm of decentralized finance, the emergence of decentralized AI (DeAI) could democratize access to AI technologies, breaking down the barriers that have traditionally favored large corporations. This could lead to a more inclusive digital economy where individuals and smaller entities can leverage AI tools and services that were previously out of reach.

Furthermore, the integration of these technologies is poised to address some of the most pressing challenges in both domains. In AI, issues like data silos and the immense computational resources required for training large models can be mitigated through blockchain’s decentralized data management and shared computational power. In the blockchain space, AI can enhance efficiency, automate decision-making processes, and improve security mechanisms. As we advance, it is crucial for developers, researchers, and stakeholders to continue exploring and harnessing the synergies between AI and blockchain. By doing so, they will not only contribute to the growth of these individual fields but also drive innovation across the digital landscape, ultimately benefiting society as a whole.

Disclaimer: This research report is exactly that — a research report. It is not intended to serve as financial advice, nor should you blindly assume that any of the information is accurate without confirming through your own research. Bitcoin, cryptocurrencies, and other digital assets are incredibly risky and nothing in this report should be considered an endorsement to buy or sell any asset. Never invest more than you are willing to lose and understand the risk that you are taking. Do your own research. All information in this report is for educational purposes only and should not be the basis for any investment decisions that you make.

General Risks for All Stablecoins

General Risks for All Stablecoins

Crypto stablecoins are a type of cryptocurrency that aims to maintain a stable value relative to a target asset, such as a fiat currency or a commodity. This is achieved through a variety of mechanisms, including fiat collateralization, algorithmic stabilization, and hybrid models. However, no stablecoin design is perfect (up to now) – each comes with its own risk, and indeed, over time we have seen events of stablecoin depeg occur, sometimes to disastrous effects for its users. In this article Reflexivity Research identifies the key risk areas underlying each stablecoin design. 

Key Takeaways

  • Given that the main function of a stablecoin is for its value to remain stable to a certain fiat curency or asset, the biggest risk is for it to display unstable prices, or even a complete depeg from the actual asset prices. The root causes of depeg events can come be due to external factors, e.g. changes to interest rates, regulatory issues, or internal factors, e.g. hacking, redemption / arbitrage risk, etc.
  • Centralized, fiat-collateralized stablecoins still command the most market share amongst stablecoins, with USDT and USDC being the most prominent examples. All of these stablecoins come with a degree of centralization risk, as they operate under the control of regulated, profit-driven entities that have the authority to mint new coins and freeze existing assets. Users also have to depend on either disclosures from the entity, or audits to have a view of the reserves backing the stablecoin. 
  • For algorithmic stablecoins, they carry the risk of all exogenous stablecoins, but also has an additional risk of endogenous collateral that is susceptible to increased reflexivity in volatile markets. The collapse of Terra’s UST stablecoin in 2020 is a prime example of this risk, as it was mainly backed by Terra’s own governance token, LUNA. With endogenous collateral, a significant decrease in value can also destabilize the stablecoin, triggering a “death spiral” for both assets.   

Price Stability / Depeg Risk 

Despite the moniker of “stablecoins,” these crypto assets are not immune to market and external pressures. They can, and do, deviate from their targeted value, usually $ 1 USD (image below). Such events are often a culmination of microeconomic factors like sudden demand surges, liquidity challenges, or alterations in the foundational collateral.

A chart simply illustrating that stablecoins rarely trade at exactly $1. Source

Macro events, like inflationary trends or interest rate amendments, can equally sway stablecoin demand. Consider an inflationary environment where the underlying asset’s purchasing power dwindles. This can compel the stablecoin to depeg. 

Externalities, such as regulatory headlines or even technical snafus like network congestion or smart contract vulnerabilities, can also be catalysts. A government’s decision to prohibit stablecoins, for instance, would severely dent demand, instigating a depegging event. 



Smart Contract Risk with the Stablecoin Protocol

Fully autonomous, smart contracts removing middlemen and operating without bias sounds great but only if the code is safe. Using stablecoins on a blockchain means users not only need to be sure the actual blockchain is safe but also the project responsible for the stablecoin. Unfortunately, that is not always the case as Level Finance experienced in 2023. Level Finance, a DeFi protocol that issues an interest-bearing stablecoin, lost $1 million due to a glitch in its smart contract.

The core of the issue lay in a flaw that permitted the nefarious attacker to produce an excessive quantity of LUSD stablecoins without the mandated collateral backup. Essentially, the attacker could mint these stablecoins at will, subsequently trading them for other digital assets, thereby illicitly extracting value from the Level Finance ecosystem.


Other Risks and Considerations

  1. Token Centralization/Governance Risk

  2. Liquidity Centralization (e.g. primarily all-in-one liquidity pool)

  3. Supply Constraints: is the growth of the stablecoin dependent on the growth of another sector (RWAs, decentralized futures, LSTs, etc.)

  4. Redemption/Arbitrage Risk: inefficient redemption mechanisms may impede the stablecoin design from functioning as designed

  5. Integrations into Other Protocols/Contagion Risk: the MIM-UST “Degenbox” Scare


Risks of Centralized, Fiat-collateralized Stablecoins

Centralization and Censorship 

In an almost paradoxical fashion, the two leading stablecoins in the crypto space, USDT and USDC, embody a significant degree of centralization and counterparty risk. Every stablecoin displays some mixture of centralized and decentralized features. However, USDT and USDC stand out as they operate under the control of regulated, profit-driven companies that possess the authoritative power to mint new coins and, quite notably, freeze existing assets. Both projects have, on dozens of occasions, “blacklisted” addresses and frozen the funds held within them, something inconceivable to the average person’s understanding of “crypto.”


Source: TheBlock

Unfortunately, many of the people who could stand to benefit from access to USDT or USDC due to their own fiat money being subjected to high inflation, instability, or capital controls are the targets of these asset freezings simply because of their home government’s relationship with the US. Citizens of countries like Iran, Syria, and Venezuela suffer from some of the highest inflation in the world (image below) but also find themselves cut off from USDT and USDC due to US sanctions against those countries. Because of this, these centralized, fiat-backed stablecoins do not pose a solution for people like them, illustrating the value in something like a truly decentralized, crypto-native stablecoin or crypto asset.


Additionally, due to their design, users of USDT and USDC must put 100% faith into Tether and Circle, respectively, to properly manage the hundreds of billions of dollars across the globe in real time. Should either company mismanage its global operation, the ramifications would be felt across the entire crypto economy.


Exogenous Collateral, Reserve Composition, and Transparent Audits

The collateral backing a protocol’s stablecoin can be categorized as either endogenous or exogenous in nature. Endogenous assets, in this context, are crypto tokens intrinsically linked to the protocol, serving roles like equity or governance (akin to LUNA in the Terra and UST ecosystem). Conversely, ETH represents an exogenous crypto asset, which is often viewed as a more desirable form of collateral because its value remains independent of the protocol’s progress and longevity. The US dollar would be ever more exogenous as it is not a crypto asset and has even less correlation to the underlying stablecoin.

Source: CoinGecko

For these fiat-backed stablecoins, the crux of investor confidence often hinges on the clarity of their financial health and operations. 

While many might assume that a “USD-collateralized” label signifies a one-to-one backing by the US dollar, the reality is more nuanced as the collateral backing these stablecoins is typically diversified into an array of assets. These might include cash equivalents such as US treasuries and commercial paper, but also encompass secured loans, and corporate bonds, among others. 

Acknowledging the importance of transparency in this domain, there has been a concerted push towards enhancing clarity and reporting mechanisms for USD-collateralized stablecoins. Moody’s, a renowned TradFi credit rating agency, is even devising a scoring methodology tailored for fiat-backed stablecoins. This system aims to assess these digital assets based on the robustness of their reserve attestations.

Circle’s USDC, one of the prominent stablecoins in the market, has bolstered its commitment to transparency, releasing monthly reports detailing its reserves with attestations from Grant Thornton, a globally recognized accounting firm.

Regrettably, when it comes to Tether, transparency into its reserve composition and internal operations poses challenges for investors and users alike. While many financial institutions and companies in today’s era leverage technology to provide real-time updates on their balance sheets, Tether is seemingly incapable of doing so. The company’s balance sheet data is updated quarterly, a cadence that might have been standard in traditional finance, but in the rapidly moving digital currency ecosystem, could be seen as outdated. This periodicity leaves a considerable gap where market movements can happen, and investors are left in the dark.

Furthermore, these quarterly reports, though informative, still leave many pertinent questions unanswered. For the astute reader, these reports lack clarity on several fronts:

  • Financial Institution Partnerships: The identity of the financial institutions that hold Tether’s cash deposits remains undisclosed. The geographical location of these institutions is equally unclear. Knowing the “who” and “where” can give investors insights into the financial stability, regulatory environment, and potential risks associated with these holdings.

  • Foreign Treasury Details: While Tether’s holdings in U.S. Treasury Bills might be known, details about non-U.S. Treasury Bills remain undisclosed. 

  • Interest Rate Exposure: The interest rate risk associated with many of Tether’s assets (and possibly some of USDC’s) is not completely known. As interest rates have exhibited extreme volatility over the last ~two years, this becomes more significant to the reserves held by each entity.

  • Loan Terms and Collaterals: Details about the terms of secured loans taken by Tether, as well as the collateral backing these loans, are missing. Such details can provide a deeper understanding of the company’s credit risk and the safety of its financial position.

Other potential hazards for USDT and USDC users include:

  • Fluctuations in the U.S. dollar value.

  • Uncertainties surrounding U.S. Treasuries.

  • Possibility of bank runs.

  • Improper risk mismanagement of funds or banking partners.

  • Internal mismanagement or even malicious practices by the overseeing company.


Banking/Counterparty/Custody Risk

Because USDT and USDC are (essentially) digital representations or IOUs of the U.S. dollar, this brings forth another layer of centralization/risk: custodying the fiat collateral that backs the stablecoin. The onus of safeguarding these funds and ensuring their seamless transfer falls upon their banking partners and all the problems/frictions involved in traditional banking. Consequently, any turbulence – operational glitches, financial insolvency, or even unethical practices – encountered by these partners can create ripples, disturbing the very foundation of these stablecoins.

Despite the risks explained above, many still feel comfortable holding USDT and USDC, acknowledging that the risks sound no different than standard banking risks. And they would be correct to a certain extent. Plus, with such a large, reputable, and regulated company behind USDC like Circle, users assume extra protections that other stablecoins can offer, like insurance. However, while users might derive a sense of security from insurance mechanisms like the Federal Deposit Insurance Corporation (FDIC), digging into the details of Circle’s recent practices illustrates how little protection users actually had in the event of a crisis. As Circle’s 2021 data suggests, out of the staggering $10 billion parked in regulated financial institutions, a mere ~$1.75 million was under the protective umbrella of FDIC (“Circle Internet Financial Limited”). In the unfortunate event of a bank collapse, USDC holders could find themselves vying with other claimants for their due. 

The reality was that a significant chunk of USDC’s backing cash was vulnerable, lying beyond the FDIC’s $250,000 insurance ceiling when it suffered banking issues in March 2023. Silicon Valley Bank ran into solvency issues and assets held at the bank were at risk. This caused USDC to depeg as ~8% of the US dollars backing the stablecoin were potentially unrecoverable. 


Regulatory Risk

Centralized stablecoins, in particular, have recently become a prominent focus for regulatory bodies, as evidenced by recent actions and proposals. One of the most noteworthy instances of this regulatory spotlight involves Paxos Trust (Paxos). The New York Department of Financial Services (NYDFS) initiated action against Paxos in 2023, instructing them to cease Binance USD (BUSD) issuance. The crux of NYDFS’s contention was Paxos’s purported negligence in its duty to carry out consistent risk assessments. 

Further intensifying the scrutiny, the Securities and Exchange Commission (SEC) advanced its allegations against Paxos, suggesting that BUSD qualifies as a security through its Wells Notice. While the immediate repercussions of these actions only pertain to BUSD, it is evident that centralized stablecoins, by extension, are vulnerable to significant regulatory interventions.

Further evidence for the growing regulatory concerns around stablecoins was the introduction of the STABLE Act in December 2020. The essence of this act is anchored in the objective of thwarting potential malfeasance, lack of transparency, and the speculative emergence of a stablecoin-driven shadow banking system.

However, the requirements proposed are seen by many as particularly onerous, to the point of potentially killing off current stablecoin designs/issuers. If the act were to be ratified, stablecoin issuers would find themselves navigating the almost impossible requirements of:

  • A necessity to procure a federal banking charter.

  • An obligation to gain endorsements from both the Federal Reserve and the Federal Deposit Insurance Corporation (FDIC), and this, a full half-year prior to any stablecoin issuance.

  • An additional stipulation to either secure FDIC insurance or establish dollar reserves directly within the Federal Reserve’s precincts.

The intersection of stablecoin innovation and regulatory oversight is becoming increasingly pronounced. For stablecoin issuers, the current unclear regulatory environment leaves them unsure of how to grow while maintaining compliance. As the crypto industry and regulators continue their discussions, the hope is for a balanced approach that ensures both security and innovation.


Oracle Dependence/Centralization

At their core, oracles compile pricing data on assets from a multitude of sources, which CDP-backed stablecoins use in their protocols’ stability mechanisms. These digital entities take on the pivotal task of ensuring the stability and reliability of the system, particularly during unpredictable market fluctuations and unforeseen black swan occurrences. The pivotal role oracles play becomes evident when considering the mechanics of crypto loans: for each loan extended, the value of the collateral must maintain a certain threshhold. Should an oracle indicate that a user’s collateral-to-loan ratio dips beneath a pre-determined mark, the collateral deposited in the associated smart contract can be mobilized and liquidated to square off the corresponding stablecoins.

Such mechanisms are integral to both Collateralized Debt Positions (CDP) and Algorithmic stablecoins. Consequently, the selection of an oracle solution isn’t a decision that should be taken lightly by the protocol. Factors that need meticulous consideration while opting for an oracle include:

  • The oracle’s capacity, reliability, and frequency in which it provides price feeds for the specific types of collateral in question, such as ETH, wstETH, or USDC.

  • Its compatibility with both Layer 1 and Layer 2 Networks

  • The inherent transparency, security, decentralization, and accuracy of the price feeds, along with the reliability of their sources.

An oracle exploit occurs when an oracle reports inaccurate data about an event or state of the external world. This can happen because the oracle purposefully acts maliciously or negligently, or the oracle’s data source is compromised.

There are two main types of oracle exploits:

  • Misreporting: This is when an oracle reports a price that differs from the correct market-wide price of an asset. Regardless of whether misreporting occurs due to malicious or negligent behavior, any protocol relying on a faulty oracle for price data may be at risk of an exploit.

  • Poor market coverage: This is when an oracle relies on only a subset of all trading environments to report the price of an asset. This can lead to the oracle misreporting the price of an asset if that subset is manipulated, even when the majority of trading environments and the market-wide price remain unaffected.

The risk of oracle exploits can be mitigated with a more secure oracle design. Features of secure oracles include:

  • Sourcing price data from across all trading environments to provide proper market-wide coverage: This helps to ensure that the oracle is not vulnerable to price manipulation in a single trading environment.

  • Protections from external tampering: This can be achieved by decentralizing the oracle or by using other security measures to prevent unauthorized access to the oracle’s data.

  • Economic incentives to report faithfully: This can be achieved by rewarding oracles for reporting accurate data and penalizing them for reporting inaccurate data.


RWA/Exogenous Risk

As RWA adoption grows in the crypto ecosystem, so too do the risks associated with them, even for CDP-type stables. Maker and Frax are both leading “crypto native” stablecoins that have begun pivoting to incorporating more RWA into their accepted collateral in the name of peg stability and reserve diversification. Essentially every risk discussed above in the “Risks of Centralized, Fiat-collateralized Stablecoins” section now applies to these RWA portions of the collateral. Anytime a protocol introduces anything non-native to its blockchain, it then inherits all the risks associated with the “real world”: counterparty, custody, regulatory, etc. Blockchain code has no ability to recognize these things and, should any issues arise, they must be handled at the social layer, diminishing the “law is code” and immutability aspects of a blockchain. To make matters worse, there is very little case law for such matters in 2023, making any legal issue involving RWAs and blockchains a new, potentially messy, and almost certainly a controversial undertaking. 



Risks of Algorithmic Stablecoins

Nearly all of the aforementioned risks above also apply to algorithmic stablecoins. However, algorithmic stablecoins contain one, extremely important, risk that the others do not: endogenous collateral that is susceptible to increased reflexivity in volatile markets. Terra’s UST stablecoin is the poster child for such risk.

UST’s “Death Spiral”/Reflexive Liquidation Risk

At its height, UST, Terra’s stablecoin, held an impressive market capitalization of over $18.7 billion, ranking as the third-largest stablecoin globally. However, as macroeconomic challenges, including rising inflation, looming rate hikes, and a general downturn in the equity markets, emerged in early May 2022, the cryptocurrency market began to decline precipitously, including LUNA, the endogenous collateral backing UST.

LUNA’s value began to crash (not uncommon in the crypto world), reducing the incentive to swap with UST, which subsequently led to the destabilization of UST’s peg to the U.S. dollar. The downward momentum on UST initiated a feedback loop, often termed a “death spiral.” Arbitrage traders exchanged UST for LUNA, subsequently selling it. This caused LUNA’s price to plummet, requiring even more LUNA to be produced for every UST redeemed. This cycle led to a rapid increase in LUNA’s supply, resulting in hyperinflation. 


As Terraform Labs intervened to restore balance by adjusting UST liquidity in key pools, their efforts proved ineffective against the prevailing selling pressure. The situation was exacerbated by a mass withdrawal from the Anchor Protocol, further devaluing UST. 

In response, the Luna Foundation Guard (LFG) liquidated significant BTC reserves, lending these assets to stabilize the UST peg. Despite these efforts, UST’s value plummeted to $0.63. Amidst the chaos, LUNA’s price experienced a sharp drop, and Terra faced insolvency concerns. Attempts to mitigate the downward spiral intensified the market sell-off, with LUNA entering a hyperinflationary state. The once-thriving Terra ecosystem witnessed dwindling confidence as all three primary liquidity pools faced disruptions, pushing LUNA’s circulating supply into hyperinflation and the UST peg into further decline toward $0.



Disclaimer: This research report is exactly that — a research report. It is not intended to serve as financial advice, nor should you blindly assume that any of the information is accurate without confirming through your own research. Bitcoin, cryptocurrencies, and other digital assets are incredibly risky and nothing in this report should be considered an endorsement to buy or sell any asset. Never invest more than you are willing to lose and understand the risk that you are taking. Do your own research. All information in this report is for educational purposes only and should not be the basis for any investment decisions that you make.