A candidate AI policy tells applicants which tools and assistance are allowed at each hiring stage. Its purpose is to make the assessment understandable before it starts. A policy should distinguish legitimate preparation, an independent interview and a task that deliberately evaluates work with AI.
Use a rule for each stage
An employer can allow AI for preparation, require independent answers during a particular interview and permit selected tools for a practical exercise. Those choices measure different things. Make the choice explicit in the invitation and repeat it at the relevant stage.
Anthropic’s candidate guidance provides one employer example: it distinguishes preparation from live interview assistance and states when exceptions are allowed. Your policy should reflect your own role and process, not copy another employer’s requirements without review.
A checklist to adapt before sending an invitation
Purpose: the role and skill being assessed at this stage.
Allowed resources: notes, documentation, calculators, AI tools and any restrictions on their use.
Independent work: whether the candidate must answer without another person or unapproved tool.
Disclosure: what the candidate should say about tool use and when.
Checks: which identity, device or session checks will take place and what preparation is needed.
Support: who to contact for technical problems or an accessibility adjustment.
Review: how a disputed observation will be checked and how the candidate can clarify it.
Information handling: the applicable notice about recording, access and retention, provided through the employer’s approved process.
This is a drafting checklist, not legal text or a claim that a single policy meets every jurisdiction’s requirements. Complete the details before using it. Do not promise a support or appeal channel that your team has not assigned.
Example: two clearly separated assessment stages
In an independent stage, a candidate explains a past decision and works through a new scenario without external answer generation. In an AI-enabled stage, the same role might require evaluating an assistant’s suggested solution and identifying what should be changed. The second stage assesses judgement when using a tool; it should not be scored as if no tool was involved.
This is a suggested assessment design. It does not imply that TrustHR provides a built-in candidate copilot or a particular switch between these modes. Confirm the workflow and available configuration with the product team.
Respond to uncertainty consistently
If an observation conflicts with the policy, record the stage, the rule and the evidence. Check the candidate’s explanation and possible technical causes. A reviewer can resolve the concern or request a comparable reassessment. Avoid inventing a new rule after the interview or treating every automated flag as a violation.
Use the same review process for similar cases. Keep a record of dismissed concerns as well as confirmed ones: otherwise a pilot can make a noisy check look effective by counting only its alerts.
How TrustHR supports the process
TrustHR combines AI interviews, generation of questions, AI scoring and TrustExam’s control capabilities. These features can be placed within an employer’s agreed process; the policy defines what conduct the process is meant to assess. The employer remains responsible for the hiring decision.
Read the AI interview questions and scoring guide when defining the assessment. Discuss a TrustHR pilot to match the checks to your candidate experience and review procedure.