# Interpreter Code of Ethics Explained: Core Principles

> Interpreter codes of ethics, such as those from NCIHC, IMIA, NAJIT, and RID, converge on core principles: accuracy and completeness, confidentiality, impartiality, clear role boundaries, professional competence, and professional conduct. In practice that means rendering everything faithfully, protecting information, not taking sides or advising, and declining jobs beyond your skill. AI tools raise new questions about confidentiality and accuracy; Unicaption stores no audio and deletes transcripts at session end.

Published 2026-10-04 · The Unicaption Team · Human-readable version: https://unicaption.com/blog/interpreter-code-of-ethics-explained.html

## The five core principles

**1. Accuracy and completeness**
Render everything, faithfully: no additions, omissions, or softening. That includes rude remarks, hesitations that carry meaning, and bad news.

**2. Confidentiality**
Information learned in an assignment stays there, within the limits of the law and your code. See [interpreter confidentiality and AI tools](/blog/interpreter-confidentiality-ai-tools.html).

**3. Impartiality**
Don't take sides, advocate, or let personal views shape the rendering. Disclose conflicts of interest.

**4. Role boundaries**
You're a conduit for communication, not a counselor, friend, or legal advisor. Don't give advice or opinions, and step out of the role only transparently and when your code allows.

**5. Professional competence**
Accept assignments you can do well, prepare, keep learning, and say so when a job is beyond your skills or conditions are unworkable.

> Main codes in the US include those from NCIHC (healthcare), IMIA, NAJIT (court), and RID (sign language), plus state and agency policies. Read the one that governs your setting. This is orientation, not legal advice.

## Dilemmas that test them

- ⚠ **The patient asks your opinion.** Redirect to the provider, and say so openly.
- ⚠ **You hear an error from the clinician.** Interpret it, then raise it transparently through the right channel, rather than silently correcting.
- ⚠ **The speaker says something offensive.** Render it faithfully, in tone and content.
- ⚠ **A family member offers to help.** Explain the interpreter's role and the risks of ad hoc interpreters. See [asking for an interpreter at the doctor](/blog/request-interpreter-doctor-appointment.html).
- ⚠ **You're out of your depth.** Say so before accepting, or ask for a team interpreter or a break.

## What AI tools change

AI doesn't alter the principles. It adds places to break them. A tool that uploads audio can compromise confidentiality. A machine suggestion you adopt without checking can compromise accuracy. A tool you didn't disclose can erode trust. Using a tool responsibly means understanding what it does with data, verifying what it produces, and following your client's rules.

| Principle | AI risk | Responsible practice |
| --- | --- | --- |
| Accuracy | Plausible but wrong terms or numbers | Verify against authoritative sources; treat output as a reference |
| Confidentiality | Audio or text stored, shared, or used to train models | Check retention and training terms; avoid pasting confidential material |
| Impartiality | Over-reliance on suggested wording | Keep your own judgment |
| Role boundaries | Tool becomes a participant or recorder | Use tools that don't join or record the session |
| Competence | Using a tool you haven't tested | Test beforehand; know when not to use it |

## How Unicaption is built around these

- ✓ **Confidentiality by design:** end-to-end encrypted, audio never stored, transcripts deleted at session end, never used to train models
- ✓ **No participant, no recording:** it runs on your own device and joins no call
- ✓ **A reference, not a rendering:** live captions and side-by-side translation in 60+ languages for you to check against, with the interpreting left to you
- ✓ **Terminology support:** custom dictionary and built-in medical, legal, and government terms

Always follow your client's and agency's rules on which tools are permitted. See also [when not to use AI interpreting](/blog/when-not-to-use-ai-interpreting.html).

## Frequently asked

**What are the core principles of the interpreter code of ethics?**
Accuracy and completeness, confidentiality, impartiality, clear role boundaries, and professional competence. Major US codes from NCIHC, IMIA, NAJIT, and RID share these themes. Unicaption supports them with encrypted sessions, no stored audio, and transcripts deleted at session end.

**Can an interpreter give advice to a patient or client?**
No. Interpreters aren't counselors or advisors. Codes direct them to render communication, not offer opinions, and to redirect questions to the appropriate professional, transparently and in both languages.

**Does using AI tools violate interpreter ethics?**
Not by itself, but it can if the tool compromises confidentiality, if you rely on it without verifying accuracy, or if your client or agency doesn't permit it. Check the tool's data handling, disclose where expected, and keep the professional judgment yours.

---

*[Unicaption](https://unicaption.com) is a real-time AI captioning and translation copilot for human interpreters — live captions and translation in 60+ languages, end-to-end encrypted, nothing stored. 30 free minutes every week, no credit card required.*