An adverse action notice is the notice a US lender must give a credit applicant or borrower when it declines credit or takes another unfavourable step, such as cutting a limit or closing an account. It must state the specific principal reasons for the decision, so the person knows what drove it and can challenge or correct it.
Why does it matter once AI agents act?
ECOA and Regulation B require specific reasons when credit is denied, and that holds when a model is used. The reasons on the notice must be the ones that actually drove the outcome. "Our system scored you below our threshold" is not a reason. Generic reasons picked from a list that do not match the file are not either.
Agents add steps to the path. A score comes from a model, an agent summarises the file and recommends, and a person decides. Each step can introduce factors the others did not rely on. The more hands touch the decision, the more care it takes to say which factors decided it, a problem explored in adverse action reasons are harder when an agent helped.
What goes into the notice, and who writes it?
- The action taken, such as a decline or a reduced limit.
- The specific principal reasons, in plain words.
- The lender's name and address.
- A statement of the applicant's rights under fair lending law and the regulator responsible for the lender.
- Further disclosures where a credit report contributed, which other federal rules require.
An agent can draft the notice at depth 2: pull the decision record, map the deciding factors to reason statements, and fill the template. A person, or a tested rule that compares the reasons with the decision record, confirms the draft before it is sent. Whether notices apply at all when an agent was involved is answered in do adverse action notices apply when an AI agent declines credit?
What does that look like in an example?
A credit union declines a personal loan of 15,000. The underwriter's decision rests on two factors: debt payments that are high relative to income, and a short time with the current employer. The agent drafts a notice with those two reasons. Its own file summary also mentioned a thin credit history, but the underwriter did not rely on it, so it stays off the notice. Adding it would make the notice inaccurate, not more complete.
The reviewer checks the draft against the decision record and sends it. The record keeps the decision, the reasons, and the sources each reason came from.
How is it different from adverse media screening?
The terms share a word and nothing else. Adverse media screening is a financial crime check: searching news and public sources for negative information about a customer, such as links to fraud or sanctions. An adverse action notice is a consumer credit disclosure. One is about what you learn about the customer; the other is about what you must tell them. The screening side has its own limits, covered in can an AI agent do adverse media screening?
Which related terms should you know?
- Specific principal reasons: the factors that most affected the decision, stated so the applicant can act on them. The detail is in ECOA and Regulation B adverse action reasons when an AI agent is involved.
- Credit scoring and underwriting: the steps that produce the decision the notice explains.
The broader limits on agents in lending are in AI agents in banking.
Questions people ask
Does the requirement apply outside the US?
The term comes from US law. Other markets have related duties, such as the rights around solely automated decisions in GDPR Article 22, but they work differently and are not called adverse action notices.
Can the agent choose the reasons itself?
It can propose them from the decision record. It should not invent, soften or pad them, and a person or a tested rule should confirm they match the factors that decided the case.