The world's leading professional interpreter association just submitted a formal statement to the Global AI Summit. Here's what they said- and why every interpreter- Deaf community member- and access advocate needs to read it.

Who Is AIIC and Why Does This Statement Matter?

The International Association of Conference Interpreters (AIIC) is one of the most established interpreter credentialing bodies in the world. In November 2024- they submitted a formal statement to the Global AI Summit — a gathering that shapes international AI policy — calling for urgent standards around AI interpreting tools.

This isn't a think piece. It's a policy intervention. And the fact that it had to be written at all tells us something important about where the industry is heading.

Two Very Different Tools- and Why the Distinction Matters

The statement opens with a distinction that often gets flattened in public conversation: Automated Speech Translation (AST) is not the same as Computer-Assisted Interpretation (CAI).

CAI tools are built for interpreters — glossary lookup- note support- feed management. They keep the human in the loop and require their own standards for transparency and data privacy. AST tools replace the interpreter entirely. The end user is the general public — and that public is- as AIIC puts it- "often overestimating the reliability of AI outputs."

When an agency says they're "using AI for interpretation-" which one do they mean? That question alone is worth bringing to your next client consultation.

The Problems AI Interpreting Has Not Solved

AIIC is direct about the current state of AST technology. Developers are still grappling with:

Hallucination — the model confidently outputs something that was never said
Catastrophic forgetting — performance degrades unpredictably on content outside training data
Racial and gender bias — baked into the data sets- often unacknowledged
Language inequity — tools perform significantly worse in lower-resource languages- meaning marginalized language communities bear the most risk

And here is the part that should give everyone pause: unlike human interpreters- who actively work to identify and address entrenched biases- AI reinforces them. This isn't a bug that will be patched in the next update. It's a structural characteristic of how these models are built.

What Standards AIIC Is Calling For

Rather than simply opposing AST tools- AIIC's statement outlines specific- actionable standards they're calling on policymakers to adopt:

Transparency of training data- including consent- copyright- and data privacy disclosures
Mandatory proficiency disclosure per language pair- with severity ratings for the types of errors that may occur
Standardized quality scoring- covering completeness- accuracy- and understandability so consumers can compare tools fairly
Clear labeling of AI-generated vs. human-produced interpretation
Confidence indicators in output- such as color-coded captions or probability percentages where the tool has low certainty

Read that list again. None of these standards currently exist in a consistent- enforceable way. You can deploy an AST tool at a medical appointment or legal proceeding today with no requirement to disclose its error rate- its language limitations- or that it's AI at all.

The statement is unambiguous on one point: the use of AI interpreting tools must remain voluntary- and individuals must always retain the right to choose a human interpreter.

This matters enormously in Deaf community contexts. Language access is a civil right- not a feature to be optimized. If a Deaf patient at a hospital is offered an AI captioning or interpretation tool because it's cheaper- do they know they can decline? Do they know what they're giving up if they accept? Are they given enough information to make that choice meaningfully?

AIIC's position is that without well-informed consent- AI interpreting should not be deployed- period.

High Risk- High Stakes

One of the statement's key policy asks is that interpreting be classified as a "high risk" activity under frameworks like the EU AI Act. This would trigger stricter oversight- mandatory testing- and liability accountability for AST tools used in sensitive settings — medical- legal- educational- governmental.

This classification doesn't exist uniformly yet. And in the United States- there's no equivalent federal framework at all.

What This Means for Interpreters Right Now

The statement ends with an appeal: AIIC and its members — practitioners- researchers- and consultant interpreters — are asking to be actively included in AI policy deliberations. Not consulted after the fact. Included in the process.

That's the same ask we need to be making in our own communities:

When your agency adopts a new AI tool- ask: Were interpreters involved in evaluating it?
When a client asks about AI captioning as an alternative to interpretation- ask: Do they understand the difference?
When policymakers talk about "AI-powered language access-" ask: Whose languages? Whose accuracy standards? Who is liable when it goes wrong?

The AIIC statement gives us language and leverage. The question is whether we use it.

This post draws on the AIIC Statement on AI for the Global AI Summit (November 2024)- submitted by AIIC France and the AIIC Science Hub AI Workstream.