Skip to main content
Featured AnalysisPrimary topicSecurity

OpenAI and Paradigm launch EVMbench to test AI on smart contract security

EVMbench evaluates AI agents' ability to detect, exploit and patch Ethereum smart contract vulnerabilities; OpenAI commits $10M to related cybersecurity research.

Feb 18, 20269:11 PMNewsroom AI

OpenAI has launched a new benchmarking system called EVMbench, developed with crypto research firm Paradigm, to test how AI agents perform on smart contract security tasks such as detecting, exploiting and patching vulnerabilities in Ethereum Virtual Machine (EVM) contracts [1][3][4].

The benchmark measures AI systems' ability to identify high‑severity bugs, generate exploit proofs, and propose patches, framing the work as a response to the financial risks posed by flawed smart contracts that routinely guard significant assets [3][2].

OpenAI said it is committing $10 million to cybersecurity research to support the benchmark and related efforts, aiming to accelerate evaluation and improvement of AI-driven security tools for the crypto ecosystem [1].

Coverage describes EVMbench as a step toward determining whether modern AI can help prevent smart contract failures and provide measurable benchmarks for researchers and developers working on crypto security [3][4].

Was this useful?

Anonymous signal used only for weekly cluster rankings. No public counters.

Share

Broadcast this coverage

Copy-ready links for the networks your audience checks first.

Support independent reporting

If this summary helped, a small tip helps keep ClusterWire running.

Privacy note: we log tip UI events (page + action, and article slug when applicable) to improve the feature. We don’t store IP address, user-agent, or wallet addresses in analytics. Tips are on-chain, so the sending address is public in the transaction.

Source Ledger

Citations

Follow the primary reporting behind this analysis. Click a citation to open the referenced source in a new tab.

  1. 3
    OpenAI and Paradigm Introduce 'EVMbench' for AI Agent Benchmarking
    Bankless News, Research and AnalysisFeb 18, 2026

Themes

Themes driving this story

Curated from the cluster of sources powering this article.

EthereumThemeSecurity/HacksThemeDeFiThemeMacroeconomy/MarketsThemeStablecoinsTheme
Live Wire

Latest Coverage

Real-time crypto intelligence ordered by publication time.