Update (25 July, 2025 at 3:00 PM UTC): This article has been updated to add commentary by Shantanu Sontakke, security engineer at web3 security firm Zokyo
Smart-contract developers and auditors told Cointelegraph that artificial intelligence usage in programming will make crypto safer despite earlier reports suggesting that AI-generated code may introduce new security risks.
A November 2024 report by the Center for Security and Emerging Technology warned that AI-assisted programming can be detrimental for cybersecurity. The paper suggests that AI generates insecure code, and models themselves are vulnerable to attacks and manipulation.
The researchers also noted that generated code may have downstream impacts by ending up in codebases that future AIs are trained on. “Our evaluation results show that almost half of the code snippets produced by these five different models contain bugs,” they stated.
A separate study in July 2024 developed a 180-task benchmark spanning 44 vulnerability types and found that the state-of-the-art AIs had a median secure-code pass rate of under 35%. Despite this, crypto smart contract developers and auditors suggest that AI-assisted coding will lead to a safer crypto ecosystem.
Related: Crypto hacks surpass $3.1B in 2025 as access flaws persist: Hacken
AI augments, but doesn’t replace developers
Developers interviewed by Cointelegraph said AI tools are being used to supplement, not replace, programmers.
0xAw, the pseudonymous lead developer at Base decentralized exchange Alien Base, told Cointelegraph that he uses AI “mostly as a tool to get quick reference checks on how to do something.”
0xAw said that he uses AI to produce “cookie-cutter” code. He recognized that with recent models, he has started “up to the idea of letting it do quick sanity checks on the code,” he said. “It now works for the really obvious stuff.”
Anton Holovchenko, a senior blockchain developer at Hacken, told Cointelegraph that he uses Cursor, an integrated development environment with AI features, for his work. He uses it for auto-completions and templating to “explain which type of contract replacement you need, and that’s it.”
Still, Holovchenko noted that the programmer should adjust the code for their needs and fix AI mistakes. He highlighted that he uses AI for templating, not vibe coding — a new term often referring to letting AI code independently, with limited human oversight.
Shantanu Sontakke, security engineer at web3 security firm Zokyo, told Cointelegraph that he used AI for programming his blockchain and AI side projects, primarily for prototyping rather than building complete smart contracts.
Related: Microsoft publishes emergency patches for its SharePoint software amid attacks
AI is common in smart contract programming
0xAw said that everybody he knows used AI for programming “to some extent, but usually as an assistant tool.”
He said, AI tools are great time savers for senior programmers, but are usually “too imprecise to do all the work themselves.”
Mike Tiutin, blockchain architect and chief technology officer at crypto forensics company AMLBot, told Cointelegraph that the prevalence of AI use is “quite high.”
Holovchenko said that “almost every developer” he knows uses some AI tools, with most relying on ChatGPT and Cursor. Sontakke noted that using AI-assisted programming environments such as Cursor is “getting quite popular” and recognized that it increases productivity and saves time.
This allows auditors to “focus more on fully exploring the attack surface and more critical and complex bugs,” Sontakke added.
AI is expected to improve Web3 security
Holovchenko added that he believes the prevalence of AI-assisted programming will lead to higher-quality code when in the hands of competent programmers:
“If you’re just vibe coding or delegating too much to AI, you will just see in the end of AI answer ‘Your code is safe now,’ and you will believe it because you don’t know how to code and your code can be actually not safe. In my opinion AI can increase the code quality but only in pair with a human who understands the coding.“
Sontakke told Cointelegraph producing documentation and code commenting has become easier thanks to AI tools. “I feel that more projects are using it to write more precise and informative documentation,” he said.
0xAw said he has noticed a trend of “junior devs using AI and thinking it can do everything for them” — often with bad results. Still, he believes that AI usage in Web3 programming is “fine” since “smart contract development requires so much testing and refinement” that bugs end up being corrected before deployment.
Instead, he raised concerns that junior developers might be unable to progress to senior roles if they excessively rely on AI tools.
Tiutin believes that AI tools will lead to “more protocols with stronger baseline security — meaning fewer careless mistakes.” He expects AI to lead to cost optimization in smart contract auditing companies without affecting the price, since what is being sold is the auditing firm’s reputation, not the audit itself.
“Audits are more about sharing responsibility than actual contract security,” he said.
Magazine: Outrage as $1.8B ‘DGCX’ crypto scam ringleader mocks victims: Asia Express