
TRACE (Threat Research and Assessment) is an AI evaluations lab dedicated to safeguarding societal systems by evaluating AI capabilities in multilingual, societal, and plural contexts. They aim to foster trustworthy AI systems by addressing complex challenges faced by open societies. Their services include evaluating AI agents and research projects, offering solutions like instruct tuning, dataset creation, developing new evaluations, and evaluating AI systems. TRACE partners with leading AI labs to enhance multilingual safety and promote responsible AI development, offering services such as ethnographic evaluation for societal alignment, multicultural/multimodal assurance, multilingual safety for global AI applications, and novel benchmarks for assessing AI comprehension across languages. They recently conducted evaluations for OpenAI on underrepresented languages, showing significant performance improvements in languages like Hausa. Their mission is to safeguard language diversity and justice through AI, providing flexible, short-term compute solutions tailored to client needs.

TRACE (Threat Research and Assessment) is an AI evaluations lab dedicated to safeguarding societal systems by evaluating AI capabilities in multilingual, societal, and plural contexts. They aim to foster trustworthy AI systems by addressing complex challenges faced by open societies. Their services include evaluating AI agents and research projects, offering solutions like instruct tuning, dataset creation, developing new evaluations, and evaluating AI systems. TRACE partners with leading AI labs to enhance multilingual safety and promote responsible AI development, offering services such as ethnographic evaluation for societal alignment, multicultural/multimodal assurance, multilingual safety for global AI applications, and novel benchmarks for assessing AI comprehension across languages. They recently conducted evaluations for OpenAI on underrepresented languages, showing significant performance improvements in languages like Hausa. Their mission is to safeguard language diversity and justice through AI, providing flexible, short-term compute solutions tailored to client needs.