Google Cloud's development platform, Vertex AI, is introducing new capabilities to combat inaccurate information in apps and services. In addition to the existing Grounding with Google Search feature, which allows models to retrieve live information, customers will now have the option to enhance their AI results using specialized third-party datasets. These datasets, provided by companies like Moody's, MSCI, Thomson Reuters, and ZoomInfo, will be integrated into the service in the third quarter of this year. These updates aim to encourage the adoption of Google's "enterprise-ready" generative AI experiences by reducing the occurrence of misleading or inaccurate outputs. Another feature, known as "high-fidelity mode, " enables organizations to source information from their own corporate datasets rather than relying solely on Gemini's knowledge bank.

Powered by a specialized version of Gemini 1. 5 Flash, high-fidelity mode is currently available for preview through Vertex AI's Experiments tool. Moreover, Vector Search, a feature that enables users to find images based on similar graphics, is being expanded to support hybrid search. This means that vector-based searches can now be combined with text-based keyword searches to improve accuracy. Furthermore, the Grounding with Google Search feature will soon introduce a "dynamic retrieval" option, automatically determining whether information should be sourced from Gemini's established datasets or Google Search for prompts that require frequently updated resources.