Citi says customers are responsible for AI agents’ actions, even after a hack
Reported by TechCrunch, Crypto Briefing, Unite.AI and 6 more outlets9 sources ↓
Citi updated its online terms to treat actions by third-party AI agents customers authorize as customer-authorized, including activity using credentials later leaked or shared. Citi says customers are responsible for monitoring those agents.
Key points
- Citi treats third-party agents’ access and actions as authorized when customers permit account access
- The terms extend that treatment to activity using credentials later shared or leaked
- Customers are responsible for monitoring agents and their actions, subject to applicable law
- Citi’s own AI tools are excluded from the third-party agent provisions
- JPMorgan, Bank of America, PNC and Citizens have also taken steps addressing third-party agent access
Sources
- TechCrunchRead ↗More banks says customers using AI agents are on their own — even if they get hacked
- Crypto BriefingRead ↗Citibank updates terms, holds customers liable for AI agent risks
- Unite.AIRead ↗Citi Makes Customers Responsible for Third-Party AI Agent Activity
- FourWeekMBARead ↗Citi Puts AI-Agent Hack Losses on Customers; 5 Banks, $9.6T
- MezhaRead ↗Citi says customers are responsible for AI agents’ actions, even after a hack
Details
Citi’s updated Online User Agreement applies when third-party AI agents access accounts or digital services through Citi Online or Citi Mobile. Unite.AI reports that the terms define agents broadly, covering software that takes autonomous, partly autonomous or non-autonomous actions on a customer’s behalf.
Under the terms, Citi treats access and actions by an agent as authorized if a customer provides credentials or otherwise permits access. Unite.AI reports that this also covers later activity using credentials that were shared, leaked or distributed, even if the actions conflict with the customer’s instructions. Customers are responsible for monitoring agents and for errors or actions beyond the authority they were given, subject to applicable law.
Citi says customers who grant third-party agents access are responsible for activity conducted through them, TechCrunch reports. The agreement excludes agents and tools provided by Citi itself, according to Unite.AI. TechCrunch reports that JPMorgan and Bank of America announced similar policies the previous month, while PNC changed its terms the same week and Citizens warned customers about third-party AI assistants.
Written from coverage as of
Images
- TechCrunch
- Unite.AI
- AndroGuider
- TechRadar
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh