Brief IA
›

Citibank Blames Clients for Errors Made by Third-Party AI Agents

AI Tools·Tom Levy·

Citibank Blames Clients for Errors Made by Third-Party AI Agents

Citibank Blames Clients for Errors Made by Third-Party AI Agents
Key Takeaways
1Citibank assigns the client responsibility for errors made by a third-party AI agent to which they have opened their account.
2Its own AI tools do not imply additional responsibility for the client in case of issues.
3Gemini offers read-only access via Google Wallet; Anthropic launched a comparable tool in early November.
Why it matters : the banks involved view the actions of authorized agents as those of the client, even when they are erroneous.

Citibank revised its terms this week: when a customer opens their account with a third-party AI agent, the bank assigns responsibility for transactions, even erroneous ones, to the customer. JPMorgan, Bank of America, Citizens, and PNC have also adopted provisions regarding these agents.

Read-Only Access and Citibank-Specific Tools

In light of the risks associated with full financial access, some AI companies are beginning to limit the permissions offered to users. For instance, Gemini, through Google Wallet, has a feature that is limited to viewing: it allows users to review expenses or track payments without permitting purchases. Anthropic proposed a similar device in early November.

At Citibank, the distinction also pertains to the provider of the tool. The bank offers several AI services to its customers. Using these services does not expose the customer to additional liability if a problem occurs, unlike using an external agent.

La lettre du soirGratuit · 19 h

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Banks Consider Agent Actions as Authorized

In the modified terms set by these American institutions, allowing an agent to access the account equates to authorizing the actions they perform. This interpretation also covers transactions made in error. Citibank also places responsibility on the customer when the agent they granted access to is hacked.

JPMorgan and Bank of America had preceded Citibank in modifying their terms. Citizens and PNC also refuse to assume responsibility for the actions of external agents to which their customers have granted access.

Citibank Warns Against Sharing Credentials

These provisions concern practices that are already present: some users seek advice from an AI before making a financial decision, share their banking data to review their expenses, or delegate purchases to their assistant. However, banks' terms generally require that the account holder log in themselves and conduct their transactions personally.

A spokesperson for Citibank explains that the institution aims to warn its customers about the risks associated with sharing credentials and granting permissions to external tools. According to him, the recent revision of the Online User Agreement aims to clarify these risks as well as the responsibility attached to the activities of authorized agents.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.