Awarded AI uses real-time internet retrieval (referred to as web search or web grounding) to improve output accuracy and reduce large language model (LLM) hallucination. For federal and defense customers, this capability is delivered through Google Cloud’s Vertex AI and Gemini services using the “Web Grounding for Enterprise” tool, which Google offers specifically for regulated industries.

Per Google’s published documentation, this configuration processes web-grounding queries within Google Cloud’s FedRAMP-authorized environment, does not log or retain customer input data, supports VPC Service Controls, and confines machine-learning processing to United States multi-regions. Customer inputs are not used to train Google’s models without permission. The sections below detail each control and cite the primary Google Cloud and FedRAMP sources so they can be verified independently.

The Google Cloud AI services that power Awarded AI’s web grounding are FedRAMP High authorized. Google Cloud holds a FedRAMP High Provisional Authority to Operate (P-ATO) issued by the FedRAMP Board (formerly the Joint Authorization Board) covering its underlying infrastructure, and Generative AI on Vertex AI (the platform hosting the Gemini family of models) and Vertex AI Search are within the set of FedRAMP High authorized services.

Because the FedRAMP Moderate control baseline is a subset of the FedRAMP High baseline, a FedRAMP High authorized Google Cloud service may be used within a FedRAMP Moderate authorization boundary. Awarded AI’s own authorization boundary is FedRAMP Moderate; customer data processed in Awarded AI is protected at the Moderate baseline.

Standard commercial grounding (Grounding with Google Search) retains prompts, context, and output for 30 days with no opt-out. To avoid that, Awarded AI uses Google’s Web Grounding for Enterprise tool (the enterpriseWebSearch tool), which Google provides for customers in regulated sectors such as healthcare, finance, and the public sector. Its documented properties:

• Curated enterprise web index. Queries are evaluated against a Google Cloud web index that is a curated subset of Google Search selected for regulated verticals, rather than against the public commercial Search endpoint.

• Zero retention of customer data. Google’s documentation states the service does not log customer data and that no customer data is persisted. Google recommends Web Grounding for Enterprise specifically where a zero-data-retention requirement applies.

• No model training on customer inputs. Under Section 17 (“Training Restriction”) of Google’s Service Specific Terms, Google will not use customer data to train or fine-tune its models without prior permission. This applies to all managed models on the platform.

The environment relies on Google Cloud Assured Workloads to enforce compliance and residency controls:

• Machine-learning processing for Web Grounding for Enterprise occurs in United States multi-regions.

• The FedRAMP High Assured Workloads Data Boundary enforces U.S. data-residency requirements.

• Web Grounding for Enterprise supports VPC Service Controls, which are used to constrain egress and reduce the risk of unintended data movement outside the defined perimeter.

Disabling web grounding is technically possible but is not recommended. Without in-boundary retrieval, the model is limited to its static training context, which degrades output quality and freshness and increases the likelihood of inaccurate responses. Several pilot teams have identified web grounding as important to their use cases. If a legal or trade-control determination ultimately requires it, the capability can be disabled as a fallback, with the understanding that output value would be materially reduced.
​

1. Google Cloud FedRAMP compliance overview — describes the FedRAMP High P-ATO issued to Google Cloud and the relationship between the Moderate and High baselines. https://cloud.google.com/security/compliance/fedramp

2. FedRAMP High authorization announcement — Vertex AI Search and Generative AI on Vertex AI achieve FedRAMP High authorization (Google Cloud blog). https://cloud.google.com/blog/topics/public-sector/vertex-ai-search-and-generative-ai-with-gemini-achieve-fedramp-high

3. Web Grounding for Enterprise — technical documentation — confirms use of the curated web index, no logging of customer data, and VPC Service Controls support. https://cloud.google.com/vertex-ai/generative-ai/docs/grounding/web-grounding-enterprise

4. Zero data retention documentation — Google’s recommendation to use Web Grounding for Enterprise where zero data retention is required, and the Section 17 training restriction. https://cloud.google.com/vertex-ai/generative-ai/docs/vertex-ai-zero-data-retention

5. Google Cloud Service Specific Terms — Section 17 (Training Restriction) and Section 19, subsection 19k (Web Grounding for Enterprise / grounding terms). https://cloud.google.com/terms/service-terms