Anthropic's AI uncovers million-dollar flaws in smart contracts

Source: CritpoTendencia Original Title: Anthropic’s AI Uncovers Million-Dollar Flaws in Smart Contracts Original Link: A recent study by Anthropic reveals that advanced artificial intelligence models can now detect and exploit vulnerabilities in blockchain smart contracts.

Additionally, according to tests with tools like Claude Opus 4.5, Claude Sonnet 4.5, and GPT-5, these agents were able to generate simulated exploits equivalent to $4.6 million, demonstrating that risks in DeFi can become automatic, systematic, and large-scale.

AI Replicated Real Attacks and Discovered New Flaws

Anthropic’s experiment was based on a set called SCONE-bench, which gathers 405 real exploited contracts between 2020 and 2025. With this material, the AI agents successfully replicated more than half of those attacks.

According to the official report, in a second phase, 2,849 recent contracts with no known vulnerabilities were analyzed. In that group, they detected two zero-day flaws and generated profitable exploits in simulated environments.

This outcome shows that automation is no longer just a hypothetical risk. Algorithms trained with language models can analyze blockchain code, identify weak points, and produce attack scripts that, in theory, could be used to target active contracts.

Although the authors emphasize that all tests were conducted in simulations and did not affect real funds, it is evident that automated cryptoattacks are a real possibility.

Hidden Vulnerabilities in the Age of AI

This development raises a serious warning, as even audited contracts could contain undetected vulnerabilities that can now be discovered automatically.

In this scenario, it becomes urgent to adopt more aggressive security protocols. This includes deeper audits, the use of white-hat artificial intelligence capable of detecting flaws, continuous code review, and the implementation of potential automated mitigation strategies.

Additionally, this revelation highlights the duality of artificial intelligence. Tools like those developed by Anthropic can be used both to strengthen security and to exploit it, underscoring the need for ethics, regulation, and proactive defense within decentralized ecosystems.

The Structural Impact on DeFi and Smart Contracts

This event could accelerate a shift in how smart contract security is evaluated. From now on, blockchain actors will need to strengthen their standards and incorporate AI-based testing, automated attack simulations, and more dynamic audits to remain protected in an environment of increasing risks.

For the DeFi ecosystem, this also serves as a call for caution. Although it offers advantages such as transparency and disintermediation, its complexity and public exposure make it an increasingly accessible target for advanced threats.

In short, this event marks a turning point. Smart contracts will need to be designed with security as a central pillar, and audits without artificial intelligence will no longer suffice. Those who take a preventive approach will be better prepared to maintain their credibility.