Fidelity lists 6 major risks to cryptocurrency AI proxy theory

Overview

Fidelity Digital Assets identifies six risks that could weaken the investment logic of combining artificial intelligence agents with public blockchains.

Fidelity believes that these six risks may prevent AI agents from creating value for the public blockchain.

Closed technology and fintech platforms may provide agents with better performance, costs, distribution channels and compliance certainty. Although payment activities can increase the activity of blockchain, they may also direct more economic value to stablecoin issuers and payment service providers. AI accelerates software development, but it also allows attackers to discover and exploit vulnerabilities at a lower cost. Over 180 days, transactions generated 49 times revenue per dollar for Ethereum than payments.

Senior Research Analyst Max Wadington released the report on August 19. Fidelity said AI could accelerate blockchain development and generate demand for programmable financial infrastructure. However, increased agency activity may not necessarily bring lasting value to the blockchain network or its native tokens.

These six risks include: limited value from increased software production, reduced technological differentiation, competition from closed systems, low-value capture in the payment chain, growing security threats, and regulatory constraints.

Fidelity views this as a possible outcome rather than a prediction. The core question of the report is not just whether AI agents will use blockchain, but whether networks and applications can derive meaningful economic value from such activities.

AI agents may prefer closed platforms over public blockchains

Fidelity describes competition from closed systems as one of the biggest risks facing the cryptographic AI narrative. Technology companies, banks, payment networks and fintech platforms are building infrastructure that allows agents to conduct transactions in a controlled environment. These platforms may have advantages in performance, cost, user experience and regulatory clarity. They already have extensive merchant distribution networks, mature identity systems and the ability to provide credit. Public blockchains cannot be taken for granted that their accessibility and programmable clearing can overcome these advantages.

"Even if AI drives a significant increase in overall digital economic activity, there is no guarantee that public blockchain will gain a significant share of it," Waddington wrote.

Fidelity anticipates that agents may use multiple types of infrastructure. For example, an agent might use blockchain for machine payments, but rely on banks or fintech platforms to process credit, identity verification and other services. The report calls this possible outcome "multi-fi."

This competition has already emerged. Companies such as Google, MasterCard, Visa, Stripe, and Coinbase are developing proxy payment systems that cover card, banking and blockchain tracks.

Payment growth may not increase the value of native tokens

Fidelity also questioned whether higher transaction volumes could bring corresponding returns to blockchain native tokens. Proxy payments can generate huge transaction volume, but the fee revenue it brings to the underlying network is limited. stablecoin issuers and payment service providers may capture more value than the underlying blockchain. Fidelity points out that low fees and fierce competition may make agency payments economically practical, but not enough to become a major source of income for token holders.

Recent activity illustrates the difference between adoption rates and revenue. As previously reported, AI agents completed 1.4 million payments on XRP Ledger, with a network fee of approximately $280. The event demonstrated technical availability, but generated negligible fee revenue relative to transaction volume. Fidelity found that in the past 180 days, transactions generated 49 times more revenue per dollar for the Ethereum base layer than payments. Transactions can also provide maximum extractable value (MEV) to the verifier.

Therefore, the report believes that agents managing capital have stronger economic potential. Costs incurred by activities such as automated transactions, lending, and providing liquidity can be much higher than large amounts of micropayments.

AI accelerates development but weakens differentiation

AI tools can help developers write, test and deploy blockchain applications faster. Fidelity cited a study involving more than 100,000 GitHub developers that found that coding agents increased the number of submissions by 180% and production release by 30%. However, Fidelity points out that more software will not automatically create useful products. Applications still require distribution channels, liquidity, regulatory compliance and ongoing user demand. For security-critical financial software, manual supervision is also necessary.

Cheaper development may make blockchain functionality easier to replicate. When competitors can quickly copy or modify similar tools, it may be more difficult for networks to distinguish themselves through the technology itself. Fidelity believes that lasting advantages may shift to liquidity, distribution channels, security and trust. Mature networks and applications may benefit because these qualities are not as easily replicated as software features.

Security and regulation will reshape proxy adoption

Fidelity said AI reduces the cost of building software, while also reducing the cost of discovering vulnerabilities and implementing attacks. The resulting pressure can transform security from a basic requirement into a core competitive advantage. The evidence supports the dual nature of this assessment. In related reports, researchers found that AI agents found real vulnerabilities in Ethereum-related software, including a flaw later disclosed numbered CVE-2026-34219. However, human researchers still need to distinguish effective discoveries from realistic false positives.

Regulatory requirements pose another obstacle. Agencies may prefer systems that provide clear identity controls, rights management, and legal accountability. A completely license-free network may have difficulty connecting autonomous agents to regulated financial services. Markets are still testing these tradeoffs. Coinbase already allows companies to accept USDC payments from autonomous agents, while Stripe, Visa and other established payment companies are developing competitive or complementary systems.

Fidelity said investors should focus on where agents deploy capital, not just how many transactions they complete. Networks that combine mobility, strong distribution channels, security, and regulatory integration may be better able to transform AI activities into lasting economic demand.