Fidelity spells out 6 risks to crypto's AI agent narrative

Fidelity Digital Assets laid out six reasons the AI agent and blockchain story might not deliver what some expect, in a report published Aug. 19.

Senior research analyst Max Wadington wrote that artificial intelligence could speed up blockchain development and create demand for programmable money. But he warned that more agent activity does not automatically mean lasting value for networks or their native tokens.

The six risks he identified: limited value from increased software production, weaker technical differentiation, competition from closed systems, low value capture from payments, growing security threats, and regulatory constraints.

Fidelity framed them as possible outcomes, not forecasts. The central question the report asks is not simply whether AI agents will use blockchains, but whether those networks can capture meaningful economic value from the activity.

Closed systems are the biggest threat

Wadington described competition from closed, permissioned platforms as one of the largest risks to the crypto AI thesis. Technology companies, banks, payment networks and fintech platforms are building infrastructure that lets agents transact inside controlled environments.

Those platforms could offer advantages in performance, costs, user experience and regulatory clarity. They already have broad merchant networks, established identity systems and the ability to extend credit. Public blockchains cannot just assume their accessibility will overcome all that, the report said.

"Even if AI drives a substantial increase in overall digital economic activity, there is no guarantee that public blockchains will capture a meaningful share of it," Wadington wrote.

Fidelity expects agents may use several types of infrastructure at once. An agent might use a blockchain for one payment but rely on a bank or fintech platform for credit, identity checks and other services. The report calls this possible outcome "multi-fi."

The competition is already visible. Google, Mastercard, Visa, Stripe, Coinbase and other companies are developing agent payment systems across card, bank and blockchain rails.

Volume does not mean revenue

Fidelity questioned whether higher transaction counts will produce proportionate returns for native tokens. Agent payments could generate enormous volume while producing limited fee revenue for the underlying network.

Stablecoin issuers and payment service providers may capture more value than base blockchains. Low fees and strong competition could make agent payments economically useful without making them a major source of tokenholder income, Fidelity said.

Recent activity illustrates the gap between adoption and revenue. AI agents completed 1.4 million payments for approximately $280 in network fees on the XRP Ledger, as reported elsewhere. The activity demonstrated technical capacity but generated little fee income relative to its transaction count.

Fidelity found that trading produced 49 times more Ethereum base layer revenue per dollar of volume than payments over the previous 180 days. Trading can also generate maximal extractable value for validators.

That is why the report sees stronger economic potential in agents that manage capital. Automated trading, lending, borrowing and liquidity provision could create more fees than large numbers of small payments.

Coding speed is not a moat

AI tools can help developers write, test and deploy blockchain applications faster. Fidelity cited research involving more than 100,000 GitHub developers that found coding agents increased commits by as much as 180% and production releases by 30%.

But more software does not automatically mean useful products. Applications still need distribution, liquidity, regulatory compliance and sustained user demand. Human oversight also remains necessary for security-critical financial software.

Cheaper development could make blockchain features easier to reproduce. Networks may find it harder to distinguish themselves through technology when competitors can quickly copy or tweak similar tools.

Fidelity said durable advantages may shift toward liquidity, distribution, security and trust. Established networks and applications could benefit because those qualities are not as easy to reproduce as software features.

Security is a double-edged sword

AI lowers the cost of building software while also making it cheaper to identify vulnerabilities and conduct attacks. The resulting pressure could turn security from a basic requirement into a central competitive advantage, Fidelity said.

Evidence supports both sides. Researchers found that AI agents identified genuine vulnerabilities in Ethereum-related software, including a flaw later disclosed as CVE-2026-34219. But human researchers still had to separate valid findings from convincing false positives.

Regulation narrows the set

Regulatory requirements create another barrier. Institutions may favor systems offering clear identity controls, permissioning and legal accountability. Fully permissionless networks could face difficulty connecting autonomous agents with regulated financial services.

The market is still testing these tradeoffs. Coinbase has enabled businesses to accept USDC payments from autonomous agents. Stripe, Visa and other established payment companies are developing competing or complementary systems.

Fidelity said investors should watch where agents deploy capital, not just how many transactions they complete. Networks that combine liquidity, strong distribution, security and regulatory integration may be better positioned to convert AI activity into durable economic demand.

Mastercard Tests Single-Audit Stablecoin Compliance With Borderless.xyz