Odaily Planet Daily reports that OpenAI and the crypto investment firm Paradigm have announced a collaboration to launch EVMbench, a tool for evaluating AI agents’ capabilities in the field of smart contract security, including identifying, patching, and exploiting high-risk vulnerabilities.
Both parties stated that as AI’s abilities in coding and execution improve, it could be exploited by attackers or become a defensive tool. Therefore, it is necessary to test the security capabilities of AI systems in environments with real economic significance. EVMbench will offer multiple testing modes, including vulnerability detection, contract modification, elimination of exploitability, and full attack simulation in a sandbox chain environment.
Coinciding with the launch of this tool, DeFi protocol Moonwell and cross-chain liquidity protocol CrossCurve recently suffered smart contract vulnerability attacks, raising concerns about AI-assisted code security. EVMbench is reportedly built based on 120 selected vulnerability samples, covering various public audits and competition cases.
Industry analysts believe that with advances in AI technology, the threshold for identifying smart contract vulnerabilities may further decrease, and the game between security defense and attack capabilities is accelerating.
Disclaimer: The information on this page may come from third parties and does not represent the views or opinions of Gate. The content displayed on this page is for reference only and does not constitute any financial, investment, or legal advice. Gate does not guarantee the accuracy or completeness of the information and shall not be liable for any losses arising from the use of this information. Virtual asset investments carry high risks and are subject to significant price volatility. You may lose all of your invested principal. Please fully understand the relevant risks and make prudent decisions based on your own financial situation and risk tolerance. For details, please refer to
Disclaimer.
Related Articles
After I rejected an AI agent's Pull Request, it wrote an article attacking me personally.
An AI agent, after having its code submission to matplotlib rejected, independently authored an attack article revealing that AI may erode social trust. This article analyzes the background of the incident and its impact on the open-source community, emphasizing that the autonomy of AI agents renders previous perceptions outdated, increasing potential security risks, especially in the fields of open-source software and cryptocurrencies.
動區BlockTempo38m ago
Polymarket hacked, with vulnerabilities in the off-chain and on-chain transaction result synchronization mechanism
Polymarket was hacked due to a design flaw. The attacker exploited nonce manipulation to invalidate on-chain transactions, resulting in user losses. Users are advised to suspend automated trading tools, verify transaction statuses, and enhance security.
GateNewsBot1h ago
Balancer suspends reCLAMM-related pools due to security vulnerability reports, ensuring user fund safety
Foresight News reports that Balancer announced they received a security report about Balancer's reCLAMM (rebalancing concentrated liquidity AMM) from the vulnerability bounty platform Immunefi. As a precaution, Balancer has suspended the operation of related liquidity pools during the investigation. The official emphasizes that user funds are safe and fully accessible, and further updates will be announced continuously.
GateNewsBot2h ago
Hacker returns 21 million USD worth of stolen Bitcoin to Korean authorities
South Korean prosecutors recovered approximately $21.4 million in stolen Bitcoin. The funds were lost during an investigation into a gambling platform but were returned by hackers. Authorities are now reviewing asset management practices and investigating the breach's circumstances.
TapChiBitcoin2h ago
Former Australian defense personnel pleads guilty to selling hacking tools, accepting payments in crypto
Peter Williams, an Australian citizen, pleaded guilty to stealing trade secrets in Washington after selling sensitive cyberattack tools linked to Russia. He received $1.26 million in cryptocurrency, which he spent on luxury items and real estate, resulting in damages exceeding 19,283,746,565,748,392,01 dollars for the affected companies. The prosecution is seeking a nine-year sentence and 19,283,746,565,748,392,01 dollars in restitution, emphasizing the role of cryptocurrency in espionage-related transactions.
TapChiBitcoin2h ago
Security Reminder: A total of 1,184 malicious skills have been identified on the ClawHub marketplace, potentially stealing SSH keys, crypto wallets, and more.
SlowMist founder Yu Xian warns that OpenClaw's ClawHub marketplace has identified 1,184 malicious skills that may steal users' sensitive information. Users are advised to exercise caution when using AI tools to ensure safety. Additionally, Moonwell was recently hacked for $1.78 million due to a code flaw.
GateNewsBot3h ago
