Zoom enhances its AI capabilities through a federated architecture, leveraging multiple language models from firms like OpenAI and Anthropic, alongside its own specialized small language models (SLMs). This method improves efficiency by utilizing SLMs for initial tasks like summarization, with powerful large language models (LLMs) like GPT-4 for refinement. “We have a committee that can detect risks,” said CEO Eric Yuan, ensuring robust oversight. Collaborating with Nvidia, Zoom plans to incorporate advanced reasoning models into its AI Companion.