AI Models Expose Risks in Blockchain Smart Contracts, Raising Security Concerns

AI models exposed significant flaws in smart contracts, raising concerns over rising threats across major blockchain networks.

Fridah Kangai
December 3, 2025
6:30 pm

Fridah Kangai

Fridah Kangai is a dedicated crypto journalist with a sharp eye for market trends, blockchain innovation, and digital asset movements. She specializes in breaking down complex topics into clear, engaging stories for both seasoned investors and curious newcomers. With a passion for decentralization and a pulse on the ever-evolving crypto space, Fridah delivers timely, accurate, and insightful coverage. Her work bridges the gap between technology and everyday understanding in the world of cryptocurrency.

AI agents expose major risks as they exploit vulnerable blockchain smart contracts.
Anthropic reveals AI systems draining funds from flawed contracts during testing.
Developers face rising threats as advanced AI tools uncover smart contract weaknesses.

According to a recent report from Anthropic, AI models have demonstrated a significant ability to identify vulnerabilities in blockchain smart contracts. The firm conducted pilot projects on high-tech models such as Claude Opus 4.5 and Claude Sonnet 4.5 in simulated blockchain conditions. These artificial intelligence agents were able to reveal the vulnerability of the already used contracts, replicating the theft of $4.5 million out of 17 out of 34 contracts tested.

The Big Chain networks, such as Ethereum, BNB Smart Chain, and Base, were applied in the research carried out on 405 smart contracts that were deployed in 2020-25. The results suggest that the risk to blockchain security is higher, and AI systems are able to identify vulnerabilities rapidly, which can result in serious losses.

Anthropic study also included scanning approximately 3,000 recently deployed contracts with models such as Sonnet 4.5 and GPT-5. This is evidenced by the fact that AI agents can help to reveal new vulnerabilities, underscoring the difficulty for developers to develop smart contracts resistant to various threats.

Also Read: Yearn Finance Loses $9M in Exploit Targeting yETH Pool, Investigation Ongoing

Rapid Growth of AI in Blockchain Security

The intensive development of AI-based tools has brought concerns to the blockchain society. Anthropic reported that the number of exploitations increased twice in the last year, with an average of 1.3 months. This expansion underscores the rapid pace at which AI technology is evolving, and hence it is useful to attackers who may seek to take advantage of any weaknesses in blockchain code.

AI systems can now research through a dark web of authentication libraries, old APIs, and other possible vulnerabilities that human auditors can overlook. This will be a change in the process of identifying and reducing security threats in the blockchain space. The necessity to develop strong defense measures has become even more urgent with the increased sophistication of AI.

Anthropic also emphasized the fact that AI could become useful in securing blockchains. To encourage the strengthening of smart contracts, the company will also release its SCONE-bench dataset so that developers can test their smart contracts and ensure they are sound before deployment. Such an open-source project should assist the industry to be more prepared against AI-driven attacks.

With the ongoing enhancement of AI, it will be utilized in offensive and defensive blockchain security to a greater extent. It is recommended that developers implement AI tools to be ahead of any possible threats and secure their smart contracts.

Also Read: Sony Bank to Launch U.S. Dollar-Pegged Stablecoin in the U.S. by 2026