AI detects XRP Ledger bug risking 18 trillion tokens and $94B market

An artificial intelligence agent has uncovered a decade-old security flaw in the XRP Ledger (XRPL) that could have minted trillions of XRP, triggering an emergency patch that bypassed traditional governance workflows.

Security firm Veria Labs reported that the vulnerability could have allowed a malicious actor to generate roughly 18 trillion XRP through a single payment transaction. This amount equals approximately 180 times the original 100 billion token supply of the cryptocurrency.

Cayden Liao, founder of Veria, noted that the defect posed a major threat to XRP’s $94 billion market capitalization by jeopardizing its fixed supply model.

Disclosed on September 22 and remediated three days later, the issue was publicly revealed on October 9. RippleX verified that no unauthorized tokens were minted, no funds were lost, and investigators found no signs of exploitation across public networks.

AI uncovers two interconnected flaws that survived a decade of audits

The discovery was made by Veria’s AI-powered security platform, which evaluated rippled—the software powering the XRPL—and found two weaknesses that could be chained together to circumvent monetary controls.

The initial weakness involved an integer overflow inside the payment engine, where purposefully arranged trading offers caused the network to miscalculate what a buyer was supposed to pay.

Under this scenario, sellers would collect their full XRP distributions while purchasers were billed only a fraction of the proper cost. The remaining balance would effectively manufacture unbacked XRP.

A second vulnerability targeted the network’s supply-defense mechanism. Because this safeguard utilized the same flawed math, it failed to detect that fresh XRP had been generated.

Carrying out the exploit required the attacker to set up hundreds of accounts and trading offers prior to executing the payment. According to official documentation, the operation only demanded a few hundred XRP in mostly refundable reserves along with standard transaction fees.

The underlying payment-engine codebase originated in 2015, whereas the impacted supply safeguard was introduced in 2017.

This longevity is striking given the network’s extensive security assessments.

Liao pointed out that the XRPL repository had been subjected to over a dozen audits and security challenges since 2024, including a competition featuring a $550,000 prize fund. Furthermore, its bug bounty initiatives have paid out upwards of $1 million.

Despite these measures, the combined defect went unnoticed until Veria’s AI system spotted it, built a functional exploit, and proved the concept on a local network.

Engineers at RippleX independently verified the exploit, confirming that the newly created XRP could indeed be spent in subsequent transfers.

Veria was awarded a $250,000 bounty, representing the maximum payout of the program. Liao characterized it as the largest known payout for a software vulnerability identified entirely by an AI agent.

Emergency fix bypasses decade-old governance procedures

The critical nature of the discovery compelled XRPL developers to make a rare choice: release a protocol-altering patch without waiting for the network’s standard amendment mechanism.

Typically, modifications to transaction rules necessitate backing from over 80% of trusted validators across two consecutive weeks before they can go live.

However, developers concluded that adhering to this timeline would leave the system vulnerable while validators debated the update.

Because the XRPL software is open source, publishing the patch directly could have also exposed the exploit details to bad actors before the defense was active.

Instead, RippleX, the XRP Ledger Foundation, and network validators coordinated an emergency deployment that turned on the protection immediately for nodes running version 3.4.1.

The update was initially released as binaries, temporarily withholding the source code to mitigate the risk of attackers reverse-engineering the weakness during the rollout.

Documentation indicates this was the first time in over ten years that the amendment activation process was intentionally bypassed for a transaction-processing modification.

This strategy came with inherent risks. Nodes operating on different versions could disagree on the validity of an exploit transaction, potentially fracturing consensus or stalling the network.

Even so, developers decided that a brief network outage was preferable to allowing counterfeit tokens into circulation.

More than 80% of validators on the default trusted list had updated by September 25, significantly mitigating that danger.

Vet, a contributor to the XRPL Foundation, stated that the synchronized reaction preserved network security and cemented version 3.4.1 as the baseline software requirement following the activation of separate Batch-related amendments on October 9.

RippleX turns to AI and formal verification after security scare

The event has driven RippleX to reevaluate how it safeguards vital infrastructure, particularly legacy code that has endured years of traditional security reviews.

J. Ayo Akinyele, head of engineering at RippleX, admitted that the flaw highlighted the necessity of re-examining long-held assumptions regarding network architecture.

He laid out plans to scale up AI-assisted vulnerability research, enhance adversarial testing, and increase oversight of legacy modules, such as the payment engine, consensus protocols, and peer-to-peer networking.

The organization also intends to fast-track formal verification, a mathematical approach that proves whether software satisfies specific security guarantees.

Those initiatives will build upon ongoing verification work concerning the XRPL Lending Protocol and Single Asset Vault, alongside partnerships with CommonPrefix and the XRP Ledger Foundation.

Akinyele noted that artificial intelligence is drastically quickening the pace of vulnerability discovery, putting pressure on developers to catch flaws before malicious parties deploy identical tools.

Vet shared that sentiment, cautioning that more sophisticated AI models could render previously obscure weaknesses much easier to capitalize on.

The October 9 disclosure also brought about a procedural update: security alerts previously marked as resolved must now undergo retesting against release candidates prior to official closure.

Leave a Reply

Your email address will not be published. Required fields are marked *