As OpenAI faces pressure to generate revenue, it has entered the realm of smart contracts with a tool to help AI agents better find weaknesses in blockchain payment environments.
The generative AI vendor on Feb. 18 introduced EVMbench, an open source benchmark for evaluating AI agents’ ability to detect and exploit vulnerabilities within smart contracts. A smart contract is a digital agreement or code stored on a blockchain that executes when certain predetermined conditions are met. If they are, a stablecoin platform can either send payments or supply new stablecoins, a type of cryptocurrency pegged to real-world reserve assets. OpenAI introduced the new benchmark together with cryptocurrency investment firm Paradigm.
EVMbench uses 120 vulnerabilities from 40 audits as data for evaluation. It also includes vulnerability scenarios identified during the security audit of the Tempo blockchain, OpenAI said. Tempo is a blockchain built by payments vendor Stripe and Paradigm for stablecoins and global payments.
OpenAI’s introduction of EVMbench is indicative of the continued growth and popularity of stablecoins and blockchain, as well as the vendor’s efforts to generate more revenue to support its business.
According to McKinsey, the total value of issued stablecoins is expected to reach $2 trillion by 2028. Moreover, smart contracts currently secure more than $100 billion in open source cryptocurrency assets, according to OpenAI. As the stablecoin market grows, cybersecurity concerns are also rising. Therefore, AI agents that can help reduce risk could be welcome. Given the opportunities AI agents offer, the cryptocurrency and stablecoin market is a revenue-generating market that could be profitable for OpenAI, which has been burning through more cash than it generates.
A Response
So, EVMbench is one response to the financial challenges the ChatGPT maker faces, said Lian Jye Su, an analyst at Omdia, a division of Informa TechTarget.
“This pressure to generate ROI and stable revenue comes with a certain bias in the sense that this is a domain that they feel like they can get into that allows them to generate sufficient revenue to justify all the valuations and investments that they made in agentic AI,” Su said.
He added that the stablecoin market has long been a target even for public cloud service providers.
“It’s not surprising that OpenAI views this market as a very interesting market for them to get into,” Su said.
Regardless of the motivation behind OpenAI’s move into the cryptocurrency and stablecoin arena, EVMbench could be attractive because it enables other vendors interested in the market to evaluate their agents and see how they perform, Su said.
A Critical Eye
However, enterprises that are more likely to adopt AI agents for their smart contract infrastructure should still take a closer look at OpenAI’s EVMbench and explore how it performs and addresses vulnerabilities compared to internal development teams, Su said.
“At this stage, different enterprises make different value judgments,” he said. “Some enterprises trust their own team more than they trust an agentic AI solution. But in the long run, as the cybersecurity landscape becomes a lot more complex, as we … address some of the vulnerabilities, some of the attack surfaces, I think it will make a lot of sense.”
As for OpenAI, it is likely the vendor will continue to explore the stablecoin market and partner with more crypto companies. Su said.
“This is just very early days,” he said. “You can almost argue that OpenAI is testing the water a little bit. I’ll be keen to see whether the other agentic AI vendors jump on this boat as well.”
