Artificial intelligence "assists" hackers in attacking old encrypted contracts, potentially earning millions of dollars.

Gate News reports that artificial intelligence is accelerating cryptocurrency hacking attacks, enabling attackers to identify old code vulnerabilities at lower costs and higher efficiency, and steal millions of dollars. Hackers are using large language models like ChatGPT and Claude to scan thousands of smart contracts per second, targeting vulnerabilities overlooked by developers and auditors.

Crypto security experts warn that outdated and poorly maintained deployments pose the greatest risks. Gabi Urrutia, Chief Information Security Officer at Halborn, points out that AI can exploit legacy contract vulnerabilities on a large scale without creating new ones, especially in old forks, abandoned deployments, and poorly maintained vaults. Attackers can easily profit from small targets while posing a low-cost threat to the entire DeFi ecosystem.

In the past, vulnerability searches were time-consuming and expensive, and only high-value contracts justified hacker investment. Now, AI allows hackers to scan thousands of smart contracts in minutes, drastically changing the attack economy. Research from Anthropic shows that their AI agents successfully exploited 63% of vulnerabilities in 405 historical smart contracts, with a total theft potential of about $4.6 million. Additionally, AI has recently discovered new vulnerabilities in newly deployed contracts and profited at very low costs.

Security researchers have observed multiple contracts being attacked repeatedly, with patterns closely matching AI-automated attacks. The recent $26 million attack on Truebit is suspected to be AI-assisted, highlighting the vulnerability of old contracts and poorly maintained codebases.

For defense, DeFi developers need to shift from the traditional “single audit” approach to continuous resource investment in AI-based vulnerability scanning. Stephen Ajayi, CTO of Hacken, states that automated AI adversarial testing will become standard practice, similar to existing penetration testing. Octane Security has successfully used AI to identify high-risk vulnerabilities in the Ethereum ecosystem, but defenses still require improved audit tracking and logging.

Experts believe that the next few years will be the most challenging period for building a decentralized economy. Gerrit Hall of Firepan notes that most DeFi protocols cannot remain secure long-term unless developers create robust contract structures that will not be exploitable over the next decade.