What is AI Agent Liability?
AI Agent Liability: an AI term, explained in plain English.
Who pays when an AI agent acting for you makes a mistake, is tricked or is hacked: you, the company that made the agent, or the bank or shop.
AI Drafted with AI help · Reviewed by Keshav Aggarwal · Updated
In simple words
Imagine you give a friend your house key to water the plants, and they leave the door open and something is stolen. Who is to blame: you, your friend, or the lock maker?
AI agent liability asks the same question when an AI helper uses your passwords to pay, book or move money.
The technical side
AI agent liability is the question of who carries the loss when an AI agent, using a person’s login or payment details, makes an unintended purchase, misreads instructions, exceeds its permissions or leaks credentials to an attacker. Banks, shops and AI companies set the answer in their terms of service, within the limits of consumer-protection law, which in the US still restricts customer liability for unauthorised electronic transfers.
On 19 November 2026 Decrypt reported that Citi’s new online banking terms treat actions taken with credentials given to an outside AI agent as authorised by the customer, even if the agent ignores instructions, exceeds its permissions or shares the credentials, while Citi’s own agents are excluded. PNC changed its digital agreement the same week to bar giving third-party agents passwords or one-time codes, and JPMorgan and Bank of America announced similar policies in October. More on our AI news feed.
AI Agent Liability kya hai?
AI agent liability ka sawaal hai: agar aapka AI agent aapke password se galat payment kar de ya hack ho jaaye, to nuksaan kaun bharega? Nov 2026 mein Citi ne kaha ki bahar ke AI agent ko login dene par uske kaam ki zimmedari customer ki hogi, aur PNC ne agents ko password dena mana kar diya.
Related terms
- AI AgentAIAn AI system that can plan steps and use tools on its own, such as searching, clicking or sending messages, to finish a task.
- Agentic CommerceAIShopping and booking done by AI agents for people: the agent finds, compares and pays for things, such as a flight, on the user’s behalf.
- Human in the LoopAIAn AI set-up where a person checks or approves the AI’s work at the important steps, so it never makes big decisions alone.
- Prompt InjectionAIAn attack where hidden or sneaky text tricks an AI model into ignoring its instructions and doing something it should not.
- AI Kill SwitchAIA control that lets people quickly stop an AI model or agent, cutting off its output or access if it starts doing something harmful.