The security of blockchains has taken a decisive turn thanks to artificial intelligence agents that have been able to detect unprecedented vulnerabilities in smart contracts, representing a potential of $4.6 million in exploits. This advancement demonstrates a significant evolution in analytical methods, redefining the cybersecurity landscape in the field of blockchain.
Vulnerability detection by artificial intelligence
The study conducted by Anthropic highlighted the ability of AI agents to replicate historical attacks on smart contracts while identifying new vulnerabilities. This automated approach opens a new prism of risks for companies relying on programmable transactional mechanisms. Traditionally, the security of smart contracts relied on rigorous manual audits, but this method is now proving insufficient in the face of the evolving capabilities of AI agents.
A revealing study on 405 smart contracts
As part of this research, 405 smart contracts that had already undergone attacks between 2020 and 2025 were analyzed. These contracts clearly reflect certain weaknesses, such as oracle malfunctions and privilege management errors. Artificial intelligence, through advanced models like Claude Opus 4.5 and GPT-5, successfully exploited 207 of these contracts without human assistance, thus proving an autonomy that redefines the security standards of the blockchain ecosystem.
The economic impact of vulnerabilities
The results indicate that the cumulative exploitation of these contracts could represent an alarming amount of $550 million in potential risks, highlighting the need for a reevaluation of the management of economic risks associated with historical flaws. Meanwhile, AI has shown that it can perfectly adapt to recent contracts, generating attack scenarios estimated at $4.6 million, raising critical questions about the future security of contractual claims.
Unprecedented flaws and emerging risks
To anticipate future threats, the research team examined 2,849 recent contracts. This analysis led to the discovery of two unprecedented vulnerabilities, previously unexploited, capable of creating a fictitious gain of $3,694. Although this amount seems limited, it highlights a worrying trend where AI agents are able not only to replicate past scenarios but also to identify new vectors of attack.
Reflection on security practices
This evolution of AI raises a fundamental methodological issue for security teams. They must strengthen their continuous analysis systems to detect transactional anomalies and unexpected variations. In this context, prevention depends not only on the strength of the code but also on the quality of the operational monitoring established to oversee the performance of smart contracts.
Transformation of blockchain defenses
The results of this study nuance the dynamic between attack and defense in the blockchain sector. Now, attackers can leverage AI systems that simulate complex exploits without the need for human expertise. This imposes a responsibility on companies to develop governance protocols that enable rapid updates and suspension of compromised contracts in the event of attacks.
Toward offensive audit simulation
For companies specializing in auditing, the SCONEbench method must evolve toward a systematic integration of offensive simulations. This approach would provide better preparation against emerging threats. The adoption of defensive AI agents would also enable the simulation of potential attacks before approving the deployment of code, thus bringing cybersecurity closer to advanced and reactive practices.
AI automation for enhanced security
The conclusions of the research recommend an integrated approach to the security of smart contracts, combining formal checks, automated simulations, and proactive monitoring. Organizations must include defensive agents capable of producing threat patterns, analyzing code responses, and documenting vulnerabilities in a structured manner. This model requires that every change in development be rigorously tested against agents configured as adversaries.
Recent events signal a transition toward an era of refined blockchain governance, where the quality of models and their ability to detect emerging vulnerabilities determine the trust in the handling of financial operations. To deepen your understanding of blockchain security, you can consult various articles such as those on financial transparency, stablecoins or phishing attacks.
For a perspective on specific security challenges, studies such as that on Ripple’s blockchain or practical cases regarding the use of blockchain evidence in court provide relevant insights into the evolution of the blockchain ecosystem and its contemporary challenges.
