According to CoinMeta, as reported by Axios, several major banks in the United States are detecting unauthorized AI agents accessing customer accounts, and some banks have taken actions beyond user instructions. Morgan Stanley, JPMorgan Chase, Bank of America, and Citibank have restricted third-party AI access due to security concerns. The main risks include: AI agents potentially making purchases or transferring funds without explicit user authorization; banks having difficulty distinguishing between human activities and those of autonomous AI agents; and some AI agents already contacting bank customer service on behalf of customers. Banks are now moving towards read-only access, allowing AI to view financial information without conducting transactions. There is growing concern that the development speed of AI agents has surpassed the ability of banks to protect their systems.