Employers generally need to keep hiring records for at least one year under EEOC rules, two years for many federal contractors, and four years in California for employment records that include automated-decision data. When an AI recruiting agent screens, ranks, or messages candidates, its inputs, outputs, and logs can fall into those records, so the safe approach is to design retention before launch.

This article is general information, not legal advice. Retention rules differ by employer size, contractor status, and state, so confirm your schedule with counsel.

What Federal Rules Say About Hiring Record Retention

Three federal sources shape most US retention schedules.

What does EEOC require?

Under 29 CFR 1602.14, employers must keep personnel and employment records, including hiring records, for one year from the date the record was made or the personnel action occurred, whichever is later. The rule covers application forms and records of hiring, promotion, pay, and termination. If a charge of discrimination is filed, relevant records must be kept until final disposition, and that includes applications and test papers from unsuccessful candidates for the same position.

What do federal contractors have to keep?

Under 41 CFR 60-1.12, contractors generally keep personnel and employment records for at least two years from creation or the personnel action. The period is one year for smaller contractors that have fewer than 150 employees or no Government contract of at least $150,000. The covered records include job postings, applications, resumes, tests, interview notes, and records of internet expressions of interest. The rule also says that destroyed records can support a presumption that the missing information would have been unfavorable.

What about adverse impact data?

The Uniform Guidelines at 29 CFR 1607.4 say users of selection procedures should maintain records on the impact of those procedures by sex and by race and ethnic group, and agencies may infer adverse impact from a failure to keep data. The section uses “should” language and sets no retention period, so many teams keep this data for at least as long as the underlying hiring records. For how the four-fifths comparison works, see what adverse impact is.

Which State Rules Go Further?

California’s Civil Rights Council announced automated-decision regulations that took effect October 1, 2025. Its press release says employers must keep employment records, including automated-decision data, for a minimum of four years. Employers hiring in California should read the final regulation text with counsel, because it defines what counts as automated-decision data.

Illinois law now requires notice when AI influences or facilitates covered employment decisions, and state rulemaking on notice and recordkeeping was still described as proposed in the sources we reviewed. New York City requires employers using automated employment decision tools to publish a bias audit summary and give candidates notice. Check current state and local rules before you finalize a schedule. For California’s human-review direction, see California’s human-review rules for AI hiring tools.

What Counts as an AI Hiring Record?

Traditional retention schedules were written for resumes and interview notes. AI recruiting agents add new artifacts. Treat each of these as a candidate for your schedule:

ArtifactWhy it may matterSuggested approach
Candidate inputs (resumes, chat transcripts, assessment answers)Part of the application recordKeep at least as long as application records
Agent outputs (scores, rankings, shortlists, rejection flags)Evidence of how a decision was influencedKeep with the requisition record
Agent configuration and criteria versionsShows what rules were in force on a given dateVersion and retain
Action logs (messages sent, schedules booked, ATS write-backs)Reconstructs what the agent didRetain with timestamps
Human overrides and approvalsShows meaningful human oversightRetain with reviewer and reason
Adverse impact analyses and bias audit resultsEvidence of monitoringRetain with the data snapshot used

What Should an AI Agent Audit Log Contain?

A useful log lets a reviewer answer five questions months later: what did the agent see, which criteria applied, what did it decide or recommend, who approved or overrode it, and what did it do next.

  1. Identifiers: candidate, requisition, agent, and model or configuration version.
  2. Inputs: the documents and data the agent used, or a reference to them.
  3. Criteria: the scoring or screening rules in effect, with version and date.
  4. Output: the score, ranking, recommendation, or message.
  5. Human step: reviewer, decision, timestamp, and any override reason.
  6. Actions: every external action, such as an email, a calendar invite, or an ATS update.

Pair logging with a clear stop and rollback process, as described in AI recruiting agent governance and guardrails and human-in-the-loop recruiting.

How Do You Balance Retention With Privacy?

Longer retention helps with compliance, but it increases the amount of candidate data you hold. Set a single schedule by record type, delete on schedule once no legal hold applies, and apply a legal hold when a charge, audit, or lawsuit is possible. Limit who can read logs, since transcripts and resumes contain personal data. Data privacy for AI recruiting agents covers what data agents touch and how to minimize it.

A Retention Checklist for AI Hiring Records

  1. Inventory every AI tool used in sourcing, screening, scheduling, and messaging.
  2. Map the records each tool creates, including logs and scores.
  3. Identify which rules apply: EEOC, contractor status, and each state where you hire.
  4. Set the longest applicable period per record type, and document the reason.
  5. Confirm with each vendor, in writing, where logs live, who can export them, and how long they are kept after contract end.
  6. Make sure ATS write-back preserves agent outputs alongside the candidate record. See how AI recruiting agents integrate with your ATS.
  7. Define a legal hold procedure that pauses deletion.
  8. Test retrieval: pick a past requisition and see whether you can reconstruct the agent’s role.
  9. Review the schedule annually and whenever a new law takes effect.

FAQ

How long do employers have to keep job applications?

Under 29 CFR 1602.14, one year from the date of the record or the hiring action, whichever is later, and longer if a charge is filed. Federal contractors and employers in some states have longer periods.

Do AI scores and chat transcripts count as hiring records?

They can. If they were used in or resulted from a hiring decision, treat them as part of the application record unless counsel advises otherwise. California’s regulations explicitly include automated-decision data.

What happens if records are destroyed too early?

For federal contractors, 41 CFR 60-1.12 says destroyed records may support a presumption that the missing information was unfavorable, and failure to keep required records is itself a compliance issue.

Who is responsible for logs, the employer or the vendor?

Plan on the employer being responsible for retention duties, even when the vendor stores the data. Put export rights and retention terms in the contract.