People who rely on interpreting need to shape its future.

High scores don’t guarantee understanding in real situations.

Language differences are not always balanced in AI apps. Risk depends on who is communicating, what languages are involved, and the situation.

Individuals, organizations, and regulators need independent evidence to make informed decisions about harm and safety.

Slide titled “The Three Dimensions of Risk”, showing three ringed circles, each labeled with one dimension: Characteristics of Primary Communicators; Machine Learning Models and Interfaces; and Setting: Typical Conditions & Tendencies.

Steadily building the foundation for language access people can trust.

  1. 2023

    Formed

    CoSET formed at the Registry of Interpreters for the Deaf (RID) Conference in Baltimore with support from the Conference of Interpreter Trainers (CIT).

  2. 2024

    Guidance and Symposium

    Symposium on AI and Sign Language Interpreting (Brown University). Helped publish Guidance.

  3. 2025

    Practical Tools

    Co-authored a toolkit on the organization, implementation, and management of enterprise-level AI interpreting software.

  4. 2026

    Sharing the Work

    CoSET has given two dozen presentations and workshops, including to the FCC and other federal entities, interpreting and access advocates, interpreter education programs, research institutes, and professional conferences.

Learn what helps people communicate effectively.

CoSET’s living testbed will examine real-world, plurilingual communication. Members discuss proposals and come to decisions while using interpreting technologies. Together, we will learn about AI’s strengths and failings and make recommendations for enhancements, safeguards, and guardrails.

The governance system of sociocracy centers communication so that everyone gives careful attention to turn-taking and the ‘moves’ of interaction. This system provides a strong baseline for assessing the effectiveness of interpreting (by AI and by humans), including how people troubleshoot various problems.

This video explains how CoSET uses sociocracy to make decisions, share responsibility across its circles, and maintain Deaf oversight.

A shared commitment to communication people can trust.

  • Our Mission

    Our mission is to protect the integrity of communication and ensure the accessibility of understanding when languages are mediated by technology and human interpreters.

  • Our Vision

    We envision a world where all interpreting needs are met, and principal communicators are satisfied.

  • Our Purpose

    CoSET SAFE AI aims to establish a rigorous, privacy-preserving, and vendor-neutral testbed for evaluating communication technologies for languages.

Tools and learning for better language access.

Help shape a fairer future for language access.

Your generosity means CoSET can measure what is safe, accountable, fair, and effective AI for human communication.

Ways to Help

  • Financial Support

    Support research and community participation.

    Donate Now
  • Partnerships

    Organizations, technology creators, and community groups can fund or collaborate on the testbed, research, or advocacy.

    Discuss a Partnership
  • Skills and Services

    Offer interpreting, translation, research, or technical support.

    Offer Your Expertise