Can DeFi Survive the Rising Threat of AI-Powered Crypto Hacks?
The decentralized finance (DeFi) industry is facing a new and unprecedented challenge: artificial intelligence. As AI systems become more advanced, security experts are questioning whether DeFi platforms can remain secure in a world where intelligent algorithms can identify vulnerabilities in seconds.
Researchers at Anthropic recently revealed that major AI models could detect critical flaws in smart contracts at an extremely low cost. Their findings showed that more than half of the exploits recorded in 2025 could have been identified—and potentially executed—by autonomous AI agents. The average cost of analyzing previously exploited contracts was only $1.22 worth of tokens.
The concern extends beyond smart contracts. Modern AI tools can rapidly uncover weaknesses in infrastructure, governance systems, and protocol architecture, creating new risks for the entire decentralized ecosystem.
DeFi Security Under Pressure
The DeFi sector has already experienced a difficult year. According to industry data, more than a dozen protocols have been targeted since April, resulting in losses exceeding $605 million.
The wave of attacks began with a $285 million breach involving Drift Protocol, reportedly linked to a combination of malware and social engineering techniques. Other affected projects included Silo Finance, which suffered from oracle configuration issues, Aethir, which experienced an access-control vulnerability, Rhea Finance, which faced problems involving fraudulent token contracts, and Volo Vault, where compromised keys played a role in the attack.
The KelpDAO Attack Sends Shockwaves Through DeFi
The most damaging incident occurred when attackers exploited KelpDAO’s LayerZero-based reETH bridge, draining approximately $290 million. The impact spread rapidly across the DeFi ecosystem, forcing more than 30 protocols to suspend certain operations as a precaution.
One of the hardest-hit platforms was Aave, which reportedly faced up to $200 million in bad debt despite maintaining some of the strongest security standards in the industry.
The incident highlighted a growing concern within decentralized finance: a protocol's security is often dependent on the security of every connected platform. Even highly secure projects can suffer significant losses when an integrated partner becomes compromised.
Could AI Slow Institutional Adoption of Crypto?
The timing of these attacks is particularly concerning because traditional financial institutions have recently increased their interest in blockchain technology and tokenized assets.
Digital asset analyst Andrew Moss warned that major security breaches could weaken confidence among banks, asset managers, fintech firms, and payment companies exploring blockchain adoption.
While most industry observers do not expect traditional finance companies to abandon crypto entirely, the pace of tokenization projects and blockchain-based financial products could slow as firms reassess security risks.
More Large-Scale Hacks May Be Ahead
Unfortunately, many analysts believe the threat is far from over. As AI capabilities continue to improve, attackers may gain access to increasingly sophisticated tools capable of identifying vulnerabilities faster than human security teams can respond.
Market sentiment reflects these concerns. Prediction market participants currently estimate a high probability that another crypto hack exceeding $100 million will occur before the end of the year.
The Future of DeFi Security
The DeFi industry has survived multiple crises, from protocol failures to market collapses. However, the rise of AI-powered vulnerability discovery presents a new challenge that may require stronger security audits, continuous monitoring, enhanced smart contract testing, and improved cross-protocol risk management.
Whether decentralized finance can adapt quickly enough remains one of the most important questions facing the blockchain industry today.
Was Artificial Intelligence Behind April’s DeFi Exploits?
So far, there is no definitive evidence proving that AI directly identified or enabled the DeFi attacks seen in April. Most of the major incidents targeted infrastructure, governance systems, or operational weaknesses rather than vulnerabilities in smart contracts. Even so, many industry observers believe artificial intelligence may have played a role.
Following the KelpDAO exploit, Bankless co-founder Ryan Sean Adams suggested that AI could be accelerating the capabilities of cybercriminals. He argued that advanced AI tools may be giving hackers unprecedented advantages, while security defenses struggle to keep pace with the rapidly evolving threat landscape.
Similar concerns were raised by independent researcher Vadim, an early contributor to the NEAR ecosystem. According to him, many smart contract vulnerabilities have existed openly for years, but locating and exploiting them required significant time, expertise, and resources. With modern AI systems, however, the cost and effort involved in discovering these weaknesses may have fallen dramatically, making attacks faster and more accessible than ever before.
“AI collapsed the cost of code analysis. Finding exploits got 100x cheaper. Writing flawless code stayed just as expensive,” he wrote.
“Use AI to find an exploit, test it on a fork, and if it works — the risk of getting caught is near zero.”
Quantstamp founder Richard Ma tells Magazine that AI discovering exploits is a “growing problem” for the sector.
“The pace of growth has accelerated significantly over the past six months as AI-powered cybersecurity tools have become increasingly advanced,” Ma explains. “Attackers have strong financial incentives, and many operate with highly specialized teams.”
According to Ma, artificial intelligence offers a major advantage because it scales far more efficiently than human labor. Instead of relying on large teams of analysts, attackers can leverage computing power to uncover vulnerabilities quickly and potentially generate substantial returns.
He notes that AI tools such as Claude Code are widely used by developers and security researchers to identify flaws before software is released. However, the same technology can also be applied to live smart contracts that are already deployed on blockchain networks.
“Standard large language models are often capable of detecting vulnerabilities directly,” Ma says. “There are very few restrictions preventing them from being used for security analysis.”
When asked why DeFi projects are not using these tools more aggressively to strengthen their own security, Ma’s response is straightforward.
“They absolutely should be,” he says. “For now, users should exercise caution when interacting with DeFi platforms until the industry catches up.”
AI Is Becoming Exceptionally Effective at Discovering Vulnerabilities
Research conducted by researchers at Anthropic examined how leading AI models performed against 405 smart contracts that had previously been exploited. The results were striking. The models successfully identified vulnerabilities linked to approximately $4.6 million in historical exploits.
Even more concerning was the rate of improvement. Researchers found that the value of exploits AI systems could identify was increasing at an exponential pace.
“Over the past year, the exploit revenue associated with 2025 vulnerabilities doubled approximately every 1.3 months,” the research team reported. On average, it cost only $1.22 in tokens for an AI system to perform a comprehensive vulnerability scan on a smart contract.
The researchers concluded that more than half of the blockchain exploits observed in 2025 could theoretically have been executed autonomously by current-generation AI agents, despite originally being carried out by skilled human attackers.
The models used in the study were also less advanced than Anthropic’s unreleased AI system, Mythos. During internal testing, Mythos reportedly discovered thousands of previously unknown zero-day vulnerabilities, including a 27-year-old flaw in OpenBSD and a 16-year-old vulnerability in FFmpeg.
To prepare for the public release of these capabilities, Anthropic has granted early access to more than 40 major organizations, including Amazon Web Services (AWS), Apple, Google, and Microsoft. These organizations are using the technology to identify and patch critical security weaknesses before more powerful AI tools become widely available.
So far, no cryptocurrency project has reportedly been granted access to the program, although Coinbase is said to be actively seeking participation.
Specialized AI is even better at finding exploits
## AI Security Research Raises New Concerns for DeFi
In a separate study, researchers from University College London and University of Sydney evaluated the capabilities of a specialized agentic AI system known as A1. The platform equips AI agents with six dedicated tools designed to analyze smart contract behavior, test exploit strategies, and interact with real blockchain environments.
According to their mid-2025 research, the A1 system achieved a 63% success rate when tested against 23 real-world vulnerable smart contracts. During the experiments, it was able to identify and execute exploits capable of extracting approximately $9.33 million in value.
Perhaps the most alarming finding was the economic imbalance between attackers and defenders. The researchers concluded that launching AI-assisted attacks can be significantly cheaper than protecting systems against them.
“Our economic analysis highlights a concerning asymmetry,” the researchers noted. “Attackers can achieve profitability with exploits worth as little as $6,000, while defenders often require protection against losses exceeding $60,000. This raises serious questions about whether AI systems naturally provide greater advantages to attackers than to defenders.”
## The KelpDAO Breach Was Not a Smart Contract Failure
Despite widespread concern about smart contract vulnerabilities, the KelpDAO incident was reportedly caused by a different weakness. The attack targeted the RPC infrastructure operating beneath LayerZero’s Decentralized Verifier Network (DVN) rather than exploiting flaws in the smart contracts themselves.
Security researcher Ma argued that the architecture contained a critical design weakness by relying on what was effectively a single point of failure.
According to him, the DVN configuration used in the system functioned more like a lone verifier than a truly decentralized verification network. As a result, compromising a single component was enough to undermine the bridge’s security.
Further evidence that the issue may have been detectable in advance came from Zengineer, a developer at TrueNorth. He stated that an AI-assisted security review conducted nearly two weeks before the breach had already identified the LayerZero DVN bridge configuration as a significant unresolved risk.
The warning reportedly highlighted concerns about the bridge’s security model 12 days before the exploit ultimately occurred, adding to the growing debate about whether AI-powered security tools can help identify critical vulnerabilities before attackers strike.
TrueNorth’s audit on KelpDAO, using its bespoke Claude Code skill two weeks ago, did highlight the DVN configuration as a potential risk. But it noted there was an “information gap” about what the configuration actually was. So the tool was unable to flag the 1:1 setup itself as a risk.
However, it highlights how AI can potentially be used to identify and zero in on potential DeFi security gaps outside of protocol logic.
AI can help with bug hunting too
## AI-Powered Bug Detection Could Strengthen DeFi Security
While AI presents new risks to decentralized finance, it is also emerging as one of the industry's most powerful defensive tools. Security experts increasingly view AI-assisted bug hunting as a game-changing technology that can help identify vulnerabilities before malicious actors exploit them.
This week, Cosmos Labs CEO Barry Plunkett revealed that artificial intelligence has dramatically increased participation in the company’s bug bounty program. According to Plunkett, researchers using AI tools are submitting far more vulnerability reports than ever before, creating both opportunities and challenges for security teams.
“AI is transforming how bug bounty programs operate,” he explained. “Researchers equipped with AI systems are generating significantly higher volumes of both valid and invalid reports. Compared to last year, submission volume has increased by roughly 900%, with our team now reviewing between 20 and 50 reports every day.”
Data from [Immunefi](https://immunefi.com?utm_source=chatgpt.com) highlights the importance of these programs. The platform reports that 61.4% of projects discover at least one critical vulnerability during their first year of operating a bug bounty program. After five years, that figure rises to 93.3%.
On average, projects uncover two critical security flaws, although some exceptional cases have identified dozens of major vulnerabilities. One project reportedly discovered as many as 50 critical issues through its security efforts.
Financial incentives remain a major attraction for researchers. The median bug bounty reward currently stands at $20,000, while the largest payout in crypto history reached $10 million for reporting a critical vulnerability affecting the Wormhole bridge. Given that advanced AI models can now analyze smart contracts for only a few dollars—or even less—the potential return on investment for successful bug hunters has become increasingly attractive.
Meanwhile, Curve researcher Chado points to a positive long-term trend. Based on an analysis of cryptocurrency and DeFi exploits over the past five years, the proportion of attacks caused by coding errors has fallen dramatically. Vulnerabilities directly linked to code bugs reportedly dropped from 37% of exploits to less than 5% by 2024.
The findings suggest that stronger auditing practices, bug bounty programs, formal verification methods, and advanced security testing are making smart contracts significantly more resilient. As AI-powered security tools continue to evolve, they may become one of the most important defenses against the next generation of blockchain threats.
Formal verification is the difficult answer
Vadim says that in future, DeFi smart contracts will need to be formally verified before they are safe enough to use.
“Assume every contract with a vulnerability will eventually be exploited. The only real defense is formal verification — mathematically proving that the code can only do what it was designed to do, before it ever gets deployed.”
Formal verification would essentially make smart contracts unhackable. Ethereum creator Vitalik Buterin has set the ambitious task of “formally verifying everything” in Ethereum. This used to be so time consuming and difficult that it was impractical, but AI makes it an achievable goal.
“We’ve also begun actively applying artificial intelligence to generate code proofs demonstrating that the software version running Ethereum does indeed possess the characteristics it’s supposed to have,” he told the Hong Kong Web3 Carnival this week.
“We’ve made progress that was impossible two years ago. Artificial intelligence is developing rapidly, so we’re leveraging this to pursue ultimate simplicity, keeping long-term protocols as simple as possible, and preparing for the future as much as possible.”
Social engineering remains a threat
But even after all the bugs have been weeded out of smart contracts, the humans in charge will remain the vulnerable part of the system. AI can be used to manipulate them too, using deepfakes and data mining. The Drift hack required six months of social engineering just to deploy the malware.
“In these times, smart contracts that have been audited are far safer than the operations around these DeFi platforms, especially operations that have key man risk susceptible to AI social engineering attempts,” Ma says.
0 Comments