The codes differ in wording and agree on the substance. What each principle asks of you at the table, on the phone, and now when a tool is running.
Every professional interpreter code, from healthcare to court to sign language, comes down to a handful of commitments that let people trust a stranger with their words. They're short on paper and difficult in practice, because real situations pit them against each other and against the instinct to help. Ethics in interpreting isn't a list of rules to recite; it's a set of habits you apply fast, under pressure, with incomplete information. This guide covers the five core principles, the dilemmas that test them, and what AI tools change.
Render everything, faithfully: no additions, omissions, or softening. That includes rude remarks, hesitations that carry meaning, and bad news.
Information learned in an assignment stays there, within the limits of the law and your code. See interpreter confidentiality and AI tools.
Don't take sides, advocate, or let personal views shape the rendering. Disclose conflicts of interest.
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.
Accept assignments you can do well, prepare, keep learning, and say so when a job is beyond your skills or conditions are unworkable.
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 |
Always follow your client's and agency's rules on which tools are permitted. See also when not to use AI interpreting.
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.
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.
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.
Encrypted, nothing stored, no bot in the call. 30 free minutes every week.
Start Free Trial →