Google Gemini Cloud adds multimodal fine-tuning and enterprise prompt management

AI · 5 min read

Gemini Cloud’s new fine-tuning pipeline accepts mixed modalities—text, image, audio—and trains specialized models that remain isolated within customer projects. Google emphasizes secure enclaves, tenant-level data governance, and audit trails to make custom multimodal models suitable for enterprise use cases like document processing and customer support.

A central Prompt Management Console lets organizations curate, version, and rollback prompt templates used across products, with runtime safeguards to prevent prompt injection and sensitive-data leakage. Administrators can set guardrails, control fallback behaviors, and receive usage analytics across teams.

Google says the updates respond to enterprise demand for more controllable AI that respects compliance needs while enabling richer multimodal capabilities. Pricing and availability are targeted at large customers starting this quarter.