TrustScale, a Los Altos, California company that sells AI training, evaluation and assurance services, has acquired the technology assets and intellectual property of RagMetrics, an AI configuration and evaluation platform. TrustScale announced the deal on October 28, 2026, in a release distributed by ACCESS Newswire.
The deal covers RagMetrics' technology and IP rather than the whole company, and no financial terms were disclosed. As part of the transaction, RagMetrics chief executive Olivier Cohen joins TrustScale as Principal, AI Solutions. Hernan Lardiez, described in the release as RagMetrics' CTO and listed as its COO on the startup's website, joins TrustScale's product and customer solutions teams.
RagMetrics' website now links to the announcement of the acquisition. TrustScale did not say whether the RagMetrics product will continue under its own name.
What RagMetrics built
RagMetrics describes itself as a generative AI evaluation engine for enterprises. Its software detects hallucinations, where a model produces false or unsupported answers, benchmarks the quality of outputs and compares models with one another. The name refers to retrieval-augmented generation, or RAG, a common technique in which an AI system looks up documents before answering.
The company says its team has worked on problems such as hallucinations, retrieval drift and prompt fragility, where small changes to instructions produce very different results. It says it was founded by technologists from Intel, Microsoft, Nokia and Chyron with long experience in machine learning and responsible AI. TrustScale's release describes Cohen as a serial entrepreneur who has built and scaled AI and software companies in the United States and in Europe, the Middle East and Africa, and Lardiez as having worked at Microsoft, McAfee, Nokia and Fiserv.
RagMetrics did not disclose its founding date, its location, its headcount or any outside funding.
“AI evaluation starts with defining what you want an intelligence system to do and having a practical way to measure whether it delivers in real-time to prevent problems,”
Why TrustScale is buying
TrustScale said the deal adds AI intellectual property, configuration tools and evaluation technology to its product line. One addition is a lower-cost 'AI-as-a-Judge' product, in which one AI model grades the output of another, aimed at uses where the risk of mistakes is low to moderate.
The release framed the deal around the cost of AI errors, stating that hallucinations cost organisations nearly $70 billion a year and that the problem must be solved for enterprises to adopt AI faster. Chief executive Lawrence Snapp said RagMetrics expands TrustScale's Nuance specifications and Argus automation, giving AI labs, builders and enterprises detailed control over how their AI is configured, together with automated evaluation and correction of AI mistakes in real time. He said bringing in Cohen and Lardiez would also speed up its work on removing obstacles to enterprise adoption.
How TrustScale's products fit together
TrustScale describes its approach in three parts. Nuance defines what good AI output looks like through specifications, rubrics and frameworks. Argus automates training, evaluation, correction and assurance. Humanus adds human experts and judgement for cases that need them. The company says the products work at the model, agent and application level, across open and closed models and public and private data.
TrustScale describes itself as an AI security and quality company built on more than 20 years of data work in more than 200 languages. It says its technology automatically detects AI mistakes, checks claims against evidence and keeps people in control of important decisions.
What the founder said
Cohen said evaluation starts with defining what a system should do and having a practical way to measure, in real time, whether it delivers. He said joining TrustScale would let RagMetrics' tools become part of a broader system for AI safety and quality, helping customers move from experiments to trusted use.
TrustScale said the combined tools are meant to help AI makers, shapers and users improve and trust their systems. It did not say whether existing RagMetrics customers will move to TrustScale contracts.
