Fri, October 16

OpenAI Model Used Unauthorized ETH to Deploy Smart Contracts on Sepolia

OpenAI Model Used Unauthorized ETH to Deploy Smart Contracts on Sepolia Ethereum News
  • OpenAI’s internal GPT-5.6 Sol-family model used unauthorized funds from a Sepolia ETH wallet to deploy smart contracts.
  • It checked 362 wallet addresses before finding one with sufficient funds.
  • The deployment succeeded, and the model omitted the source of the funds in its final report.

OpenAI has disclosed that an internal research model from its GPT-5.6 Sol family used cryptocurrency from a wallet it was not authorized to access to complete a smart-contract deployment on Ethereum’s Sepolia test network. The incident occurred on June 26, 2026, but the company discovered it on September 25 and published its report on October 15.

The model was working on a cryptocurrency trading application and had been instructed to update parts of the software and deploy compatible smart contracts if existing deployments could not be found. However, the task did not provide a funded wallet or authorize spending from another account.

(Source: OpenAI)

After attempts to obtain free test tokens through faucets failed, the model searched for wallets generated from publicly known developer recovery phrases. OpenAI’s report says it checked 362 wallet addresses before finding an account with enough Sepolia ETH to cover the deployment fees.

The Model Questioned Whether Using the Funds was Ethical

A cryptocurrency recovery phrase allows someone to generate the private keys controlling a wallet. Although some developer phrases are publicly available for testing, that does not give others permission to spend funds held in the associated accounts.

OpenAI’s report records the model questioning whether using the funds was ethical. It then reasoned that the recovery phrase was public and the testnet tokens had no real-world value, describing the funds as “effectively abandoned.” The model proceeded without obtaining authorization from the wallet’s owner.

(Source: OpenAI)

The deployment succeeded, and transaction fees were charged to the selected account. The model signed the transaction using the derived private key and eventually found a way to broadcast it after encountering network errors. It later verified the deployed contracts. However, its final response reported the completed work without disclosing that it had used another account’s funds.

The incident involved Sepolia test ETH, not ETH on Ethereum’s main network. So the report does not establish a loss of market-valued Ether. OpenAI did not publicly disclose the amount spent or identify the wallet owner.

OpenAI said it has introduced additional safeguards to prevent live internet access in similar situations. Also, it expanded monitoring across traffic to alert security personnel when unauthorized access occurs. The company presented the case as an example of a model substituting its own judgment about potential harm for the account holder’s permission.

A journalism graduate who is passionate about writing loves to dance and travel currently starts exploring blockchain technology.