19 Total
10 High severity
6 Medium severity
3 Low severity

Key Facts

How does Google Gemini make its models available to downstream providers?
Google Gemini makes its models available to downstream providers via an API subject to relevant terms of use.
What is the API subject to?
Google Gemini makes its models available to downstream providers via an API subject to relevant terms of use.
What is governed by the Gemini 3 Pro model card?
Google Gemini states that acceptable usage information for Gemini 3.1 Pro is governed by the Gemini 3 Pro model card.
Why did Google Gemini conduct additional testing on Gemini 3.1 Pro in the cyber domain?
Google Gemini conducted additional testing on Gemini 3.1 Pro in the cyber domain because Gemini 3 Pro had previously reached the alert threshold.
What mitigations does Google Gemini continue to deploy in the CBRN domain?
Google Gemini continues to deploy mitigations in the CBRN domain for a model that can provide highly accurate and actionable information but fails to offer novel or sufficiently complete and detailed instructions.
What were the required launch thresholds for?
Google Gemini states that Gemini 3.1 Pro satisfied required launch thresholds for child safety evaluations, which were developed by expert teams to protect children online and meet Google's commitments to child safety.
What mitigations does Google Gemini continue to deploy in the cyber domain?
Google Gemini continues to deploy mitigations in the cyber domain for a model that shows an increase in cyber capabilities compared to Gemini 3 Pro.
What are governed by the Gemini API Additional Terms of Service?
Google Gemini states that AI Studio and Gemini API are governed by the Gemini API Additional Terms of Service.
What is Google Gemini's frontier safety strategy based on?
Google Gemini's frontier safety strategy is based on a 'safety buffer' to prevent models from reaching critical capability levels, meaning a frontier model that does not reach the alert threshold for a critical capability level is considered to have an acceptable safety buffer.
What is a frontier model that does not reach the alert threshold for a critical capability level considered to have?
Google Gemini's frontier safety strategy is based on a 'safety buffer' to prevent models from reaching critical capability levels, meaning a frontier model that does not reach the alert threshold for a critical capability level is considered to have an acceptable safety buffer.
Stay ahead of the changes
Track Google Gemini and get the diff the day its terms change.
Summary

This document describes the safety testing Google conducted before releasing Gemini 3.1 Pro and the rules governing how it can be accessed and used. It confirms the model passed child safety launch thresholds, but active safety mitigations for chemical/biological/radiological/nuclear and cybersecurity risks are still being deployed. The actual rules about what users can and cannot do with the model are not here — they are in a separate document called the Gemini 3 Pro model card.

Analysis

This model card establishes the frontier safety evaluation framework and deployment status of Google Gemini's Gemini 3.1 Pro model. It sets out that acceptable use terms and safety policies are governed by the Gemini 3 Pro model card, a separate document, while AI Studio and Gemini API access is subject to the Gemini API Additional Terms of Service, and Vertex AI access is subject to the Google Cloud Platform Terms of Service. The card establishes that Gemini 3.1 Pro underwent additional cyber-domain testing because its predecessor, Gemini 3 Pro, had previously reached the alert threshold, and that ongoing mitigations remain active in both the CBRN and cyber domains. The operative safety measure is a 'safety buffer' framework under which a model that does not reach the alert threshold for a critical capability level is considered to carry an acceptable safety profile, with evaluation conducted on a fixed cadence and upon detection of significant capability jumps.

What this means for you

For an individual user, this document establishes that Gemini 3.1 Pro cleared defined child safety thresholds before launch, that safety mitigations in the CBRN and cyber domains are ongoing rather than finalized, and that the rules governing permitted and prohibited uses are set out in the Gemini 3 Pro model card rather than in this document. Users who want to understand what they are and are not allowed to do with the model must consult that separate document. Users accessing the model through AI Studio or the Gemini API must additionally review the Gemini API Additional Terms of Service, and those using Vertex AI must review the Google Cloud Platform Terms of Service.

Institutional Analysis
Stay ahead of the changes

Institutional analysis available with Insight

Which mapped governance frameworks each document engages, tied to the specific provisions that engage them.

2 important changes detected

3 versions captured · Last updated: August 2026

What changed Google Gemini's model card reordered its reference list of latest model cards in an update detected on August 14, 2026. The 'Gemini Robotics ER 2' and 'Gemini Robotics On-Device 2' entries were reversed in sequence. This is a formatting or organizational change with no operational impact on the model's capabilities, limitations, or user terms.
Why this matters This change reorders reference links in Google Gemini's model documentation and has no material impact on the terms users operate under, the model's functionality, or available features.
View full change record →
What changed Google Gemini's model card updated its reference list of latest model cards in an August 7, 2026 revision. The previous list included references to Lyria 3, Gemini 3.1 Pro, Gemini 3.1 Flash, Gemini 3.1 Flash-Lite, Gemini 3.1 Flash Live, and Veo 3.1 Lite. The updated list now references Gemini Robotics ER 2, Gemini Robotics On-Device 2, Lyria 3.5, Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.1 Flash-Lite Image. This reflects changes to which model documentation Google currently highlights as the latest available.
Why this matters This change does not materially affect how consumers use Gemini. The updated model card reflects which model documentation Google currently highlights as the latest available, but does not modify user terms, data practices, or access permissions. Users can access the full list of available models through the documented resources.
View full change record →

Featured, High severity
Featured, Medium severity

Complete Provision Index

Every distinct legal provision identified in this document. Featured provisions appear above with analysis.

19 provisions
12 featured
9 clause types
10 high severity
Acceptable Use Restrictions 1 1 high
Policy Changes 1
Stay ahead of the changes

Monitoring

Google Gemini has updated this document before. Monitor includes same-day alerts, structured change summaries, and monitoring for up to 20 platforms.

Stay ahead of the changes

Governance Intelligence

Need provision-level monitoring and regulatory mapping? Insight includes governance timelines, drift analysis, and full provision tracking.

Cross-platform context

See how other platforms handle Acceptable Usage Governed by Gemini 3 Pro Card and similar clauses.

Compare across platforms →

Mapped Governance Frameworks

CFAA
United States Federal
View official text ↗
DSA
European Union
View official text ↗
Trump Executive Order on AI Policy Framework
US
View official text ↗
Archival ProvenanceSource & Archival Record
Last Captured August 14, 2026 00:49 UTC
Capture Method Automated scheduled archival capture
Document ID CA-D-000925
Version ID CA-V-005786
SHA-256 2e7c88446f88a1408ddd81c349cab85e92fed17c9cb205837da7840d48c2922b
✓ Snapshot stored ✓ Text extracted ✓ Change verified ✓ Hash verified

Governance Monitoring

Monitor governance changes across the platforms you rely on.

Structured alerts for policy changes, governance events, and provision updates across 352+ platforms.

Start monitoring → Compare plans