JPMorgan, BofA, Citi & Other US Banks Restrict AI Agents; CZ, DeSantis, Sanders Flag Risks
Major US banks are restricting third-party artificial intelligence (AI) agents from accessing customer accounts through shared login credentials, following concerns that some automated tools may attempt actions beyond users’ instructions.
An Axios report published on October 22 said banks including JPMorgan Chase, Bank of America, Morgan Stanley, Citi, and Wells Fargo were responding to the risks posed by outside AI agents. Reuters separately reported that banks had been blocking such tools, citing people familiar with the matter, while noting it could not independently verify the original report.
Banks block third‑party AI agents after discovering credential‑sharing tools could autonomously initiate transactions.
Rogue AI agents, potentially thousands, exploit login data to bypass user instructions, prompting precautionary bans.
Crypto leaders warn AI‑enhanced phishing can misuse stolen credentials without breaking blockchain cryptography.
The issue has drawn attention from the crypto industry because it raises questions about how autonomous software should be authorized to access financial accounts, what actions it should be permitted to take, and who bears responsibility when something goes wrong.
Changpeng Zhao, the Binance co-founder widely known as CZ, linked the debate to crypto security. The issue has also drawn reactions from Florida Gov. Ron DeSantis and Sen. Bernie Sanders.
Why Banks Are Restricting Third-Party AI Agents
The concern is not simply that AI systems can access financial information. It is that some tools may use a customer’s login credentials to sign in directly, potentially gaining the ability to perform actions through the account.
During JPMorgan’s third-quarter earnings call on October 13, CEO Jamie Dimon said the bank would not allow just any external agent to access its systems, citing risks to customers and the institution. He warned that “rogue agents” could number in the thousands.
JPMorgan CFO Jeremy Barnum also criticised credential sharing and screen-scraping, saying the bank was blocking outside agents where possible to protect customers and itself. He said the bank was working on ways to support these tools more securely.
JPMorgan’s position does not amount to a rejection of all AI tools or every form of third-party data access. The bank has distinguished between external agents logging in with a customer’s credentials and customer-authorized data sharing through open-banking providers.
That distinction matters: a controlled connection that permits an application to retrieve account information is different from giving an agent credentials that may let it initiate payments or other transactions.
What Other Banks Are Reporting
The response across financial institutions is not uniform. Reuters’ report stated that Morgan Stanley had blocked third-party agents from accessing its platforms, but that the firm had not seen instances of rogue agents accessing them.
Charles Schwab CEO Rick Wurster has also said the brokerage blocks unauthorized agents attempting to log in for clients. His comments reflect a broader effort to control how external software authenticates and interacts with financial accounts.
The distinction between confirmed incidents and precautionary restrictions is important. A bank blocking an agent does not necessarily mean the agent successfully accessed an account, moved funds or caused a loss.
The emerging challenge for banks is to distinguish customer-directed activity from actions an AI agent may initiate independently, while allowing legitimate, permissioned uses of automated tools to develop.
CZ Warns AI Does Not Need to Break Bitcoin’s Cryptography
The banking report prompted CZ to focus on a different security concern: AI-assisted phishing and social engineering.
In his post on X, CZ argued that AI does not need to break Bitcoin’s cryptography to cause damage. Leaked personal information, stolen credentials and convincing impersonation attempts can give attackers ways to target users without defeating the cryptographic systems that secure the blockchain.
That distinction matters because the risks are different. Breaking Bitcoin’s cryptography would involve compromising the underlying mathematical mechanisms that protect the network. Phishing and social engineering instead attempt to manipulate people into revealing sensitive information, connecting to malicious applications or approving transactions.
For crypto users, an AI-generated message impersonating an exchange, wallet provider or trusted contact could make a fraudulent request more convincing. AI may also help attackers produce or personalize messages at greater scale, although the extent to which it increases successful crypto theft should be supported by incident data.
CZ’s warning therefore points to a familiar security principle: the strength of blockchain cryptography does not protect users who disclose their seed phrases, expose private keys or approve malicious transactions.
Lawmakers Warn of Financial Risks From Autonomous AI
The reported activity has also prompted political reactions, with DeSantis warning about the potential scale of financial losses and Sanders calling for a pause on AI.
In a post on X, Florida Gov. Ron DeSantis asked readers to imagine a rogue AI agent liquidating a person’s bank account and investment portfolio. DeSantis argued that an AI company might be able to compensate an individual customer for losses, but could become insolvent if rogue agents caused losses worth hundreds of billions of dollars across millions of accounts. He said criminal liability for AI executives might help deter misconduct but would be insufficient on its own to compensate victims after major losses.
Sen. Bernie Sanders took a more direct position on the use of autonomous agents in financial accounts. He called the reported situation “insane” and said AI agents should not be allowed to operate independently inside bank accounts without authorization. Sanders warned that unchecked AI could threaten the broader financial system and called for a pause in AI development, writing: “Pause AI NOW!”
The two politicians raised distinct policy concerns. DeSantis focused on accountability and whether AI companies could cover widespread losses, while Sanders called for a broader pause in AI development.
Why the Banking Debate Matters for Crypto
The banking restrictions contrast with efforts across crypto to let AI agents access services, make payments and interact with blockchain applications under defined permissions.
For example, Coinbase has outlined tools for AI agents to interact with crypto accounts and make stablecoin payments. The company’s approach includes user-defined permissions and limits, alongside payment infrastructure designed for automated transactions.
Other projects are exploring safeguards that go beyond simply enabling transactions. XDC AI’s integration with Concordium is designed to help counterparties verify the person or business associated with an AI agent. Such identity checks can help establish accountability, but they do not prove that an agent’s actions are safe or appropriate.
These developments illustrate a key distinction: enabling autonomous payments is not the same as giving an AI system unrestricted control over a financial account. Identity verification, narrowly scoped permissions, spending limits, and transaction approvals address different parts of the security problem.
For blockchain-based systems, the fact that a transaction is recorded onchain does not guarantee that the person authorizing it understood the request or that the agent acted as intended.
Security Depends on Access Controls and User Authorization
For consumers, the reported banking activity reinforces several existing security precautions:
- Protect credentials and private keys. Do not share wallet seed phrases or private keys with AI tools or unverified applications.
- Limit permissions. Give applications access only to the information and actions they need.
- Use transaction controls. Where available, set spending limits, use separate wallets for automated activity and require approval for high-value transactions.
- Verify suspicious requests. Check unexpected messages through a trusted channel instead of relying on the message itself.
- Review connected applications. Revoke permissions that are no longer needed and monitor accounts for unexpected activity.
These measures cannot eliminate every risk, but they can reduce the potential consequences of compromised credentials, misleading instructions or an agent operating outside its intended scope.
The central issue is not whether all AI agents should be allowed or prohibited. It is how financial institutions and technology providers can establish reliable authorization, limit the actions automated tools can take and determine responsibility when those safeguards fail.
The reported restrictions are an evolving response to that challenge. The extent of any resulting financial losses, the involvement of specific AI products, and the eventual regulatory response remain separate questions that require further evidence.
Also Read: SBI Data Breach: 3,897 Customer Records on the Dark Web. Can Blockchain Protect India’s Banks?
Disclaimer: The information researched and reported by The Crypto Times is for informational purposes only and is not a substitute for professional financial advice. Investing in crypto assets involves significant risk due to market volatility. Always Do Your Own Research (DYOR) and consult with a qualified Financial Advisor before making any investment decisions.


