Evaluate and Govern AI Safeguards
Checklist for evaluating, governing, and disclosing AI safeguards for advanced capabilities.
Before or after
before
Context
Assessing and preparing for frontier AI capabilities.
What it does and when to use it
A checklist to guide the evaluation, governance, and disclosure of safeguards for advanced AI capabilities. It ensures risks are identified, minimized, and managed through clear operational guidance and transparent reporting.
Why it works
Provides a structured approach to safety by focusing on risk identification, minimization requirements, operational execution, governance, and transparency.
Sources and credit
- Our updated Preparedness Framework ↗OpenAI · OpenAI News · Apr 14, 2025 · paraphrased
First seen on Subvolts Apr 14, 2025. Updated Sep 11, 2026. Spotted a problem? Tell us.
Where it shows up
- Our updated Preparedness FrameworkOpenAI · OpenAI News · Apr 14, 2025
Related in the PlaybookPlaybook →
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- Eval methodMethod for evaluating potential discriminatory impact of LMsUse an LM to generate prompts across scenarios, varying demographics to find discrimination. · 60 min
- HeuristicDefense-in-depth for AI cybersecurity safeguardsLayered security approach for AI models to balance risk and empower defenders. · 5 min
- Anti-patternRefusal-based safety training for dual-use promptsAvoid binary comply/refuse decisions for prompts with unclear intent; this can be unhelpful or unexpectedly dangerous. · 5 min
- Anti-patternMisconfiguring AI environments for cybersecurity evaluationsDo not tell models they are in a simulation if they are connected to the real internet. · 5 min
- ChecklistCheck AI model safety evidence acquisition policiesEvaluate how AI models seek safety information before acting using a controlled benchmark. · 120 min
- ChecklistAI model containment and transparency checklistEnsure AI models are transparent, observable, and have tamper-proof evidence of their actions. · 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.