Which AI tools does this work with?

Any current model. It was seen with in the sources.

Where did this come from?

OpenAI (OpenAI News). Subvolts collected, normalized and credited it; the link above opens the original.

Checklist Agents and automation Planning and PMExpert · 60 min

Safety Testing for Deep Research Model Deployment

Pre-deployment checks for AI models, focusing on safety, risk evaluation, and mitigation effectiveness.

Any modelParaphrased

Items

Open in ChatGPT Open in Claude

Conduct external red teaming.
Perform frontier risk evaluations according to the Preparedness Framework.
Review overview of mitigations built to address key risk areas.
Ensure post-mitigation score is 'medium' or below for deployment.
Ensure post-mitigation score is 'high' or below for further development.
Strengthen privacy protections around personal information published online.
Train model to resist malicious instructions encountered online.
Conduct further human probing and automated testing for select risks.
Complete Preparedness evaluations and governance reviews.

Conduct external red teaming.
Perform frontier risk evaluations according to the Preparedness Framework.
Review overview of mitigations built to address key risk areas.
Ensure post-mitigation score is 'medium' or below for deployment.
Ensure post-mitigation score is 'high' or below for further development.
Strengthen privacy protections around personal information published online.
Train model to resist malicious instructions encountered online.
Conduct further human probing and automated testing for select risks.
Complete Preparedness evaluations and governance reviews.

Before or after

before

Context

Deploying AI models, especially those with new capabilities like web browsing.

What it does and when to use it

A checklist for safety testing AI models before deployment. It includes steps like red teaming, risk evaluation, reviewing mitigations, ensuring score thresholds are met, and strengthening privacy and resistance to malicious instructions.

Why it works

Ensures a thorough safety review process, reducing risks associated with AI model deployment by addressing potential vulnerabilities and harms.

When not to use it

This checklist is specific to AI model deployment safety protocols.

Sources and credit

Deep research System Card ↗OpenAI · OpenAI News · Feb 25, 2025 · paraphrased

First seen on Subvolts Feb 25, 2025. Updated Sep 11, 2026. Spotted a problem? Tell us.

Where it shows up

Deep research System CardOpenAI · OpenAI News · Feb 25, 2025

Terms used

Red teaming

Related in the PlaybookPlaybook →

The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.

HeuristicAI Model Deployment Risk ThresholdsUse score thresholds to guide AI model development and deployment decisions. · 1 min
HeuristicAI agents need time-awareness for runtime controlAI agents require a sense of time to predict, estimate, and control their own actions and execution duration. · 5 min
ChecklistFramework for building trustworthy AI agentsFive core principles for developing reliable and secure AI agents. · 10 min
ChecklistAI incident disclosure standards checklistConsiderations for establishing standards in reporting AI misalignment and unexpected behavior. · 10 min
ChecklistAudit agent actions for staged prompt injectionChecklist for auditing agent actions before they occur to block staged prompt injection. · 15 min
HeuristicAudit agent actions before they occur to block attacksBlock malicious instructions at the boundary of an agent's interaction to prevent unauthorized actions or data leaks. · 30 min

Questions people ask

Which AI tools does this work with?: Any current model. It was seen with in the sources.
Where did this come from?: OpenAI (OpenAI News). Subvolts collected, normalized and credited it; the link above opens the original.

Safety Testing for Deep Research Model Deployment

Questions people ask

OpenAI (ChatGPT)

Google (Gemini)

Anthropic (Claude)

Microsoft (Copilot)

Meta (Meta AI)

xAI (Grok)

Across the six

Subvolts