OpenAI Preparedness Framework
OPENAI-PREPAREDNESS-2023 · US
OpenAI Preparedness Framework is a Voluntary code from US, adopted on 2023-12-18 and effective 2023-12-18. Current status: In force. Capability-tier risk evaluation regime with four categorical levels (Low / Medium / High / Critical) across four risk categories (cybersecurity, CBRN, persuasion, model autonomy). Pre-deployment evaluation against the framework gates release decisions; Safety Advisory Group + board-level Safety & Security Committee govern threshold determinations. Seoul Frontier AI Safety Commitments signatory. NOTE (iter-314): the Preparedness Framework is a versioned-evolving artefact; this row pins the originally-published Dec 2023 version, but OpenAI publishes updates on the safety/preparedness page. Citers tracking specific risk-category language or threshold definitions should confirm against the current published version — the catalog re-pins on the next Coverage Games event.
Scope and obligations
Capability-tier risk evaluation regime with four categorical levels (Low / Medium / High / Critical) across four risk categories (cybersecurity, CBRN, persuasion, model autonomy). Pre-deployment evaluation against the framework gates release decisions; Safety Advisory Group + board-level Safety & Security Committee govern threshold determinations. Seoul Frontier AI Safety Commitments signatory. NOTE (iter-314): the Preparedness Framework is a versioned-evolving artefact; this row pins the originally-published Dec 2023 version, but OpenAI publishes updates on the safety/preparedness page. Citers tracking specific risk-category language or threshold definitions should confirm against the current published version — the catalog re-pins on the next Coverage Games event.
OpenAI Preparedness Framework addresses 3 contested AI-governance topics explicitly, 5 via general principles,.
Topics governed
- governsFoundation Models / GPAI— Preparedness Framework §1-2 — applies to all OpenAI frontier-model releases
- implicitCompute-Threshold Reporting— Capability-tier evaluations are the primary trigger; compute is a coincident signal
- implicitTransparency Obligations— Public Preparedness Reports + Safety Advisory Group decisions; full evaluation methodology partially disclosed
- governsCatastrophic & Existential Risk— Preparedness Framework risk-tier matrix — Critical tier explicitly targets CBRN, cyber, persuasion, autonomy
- implicitInternational Coordination— Seoul Frontier AI Safety Commitments signatory; pre-deployment evaluation sharing with US + UK AISIs
- governsAgentic AI Governance— Preparedness Framework — Model Autonomy is one of four named risk categories
- implicitOpen-Weight Frontier Release— Framework applies to OpenAI deployments (closed-weight); does not address third-party open release
- implicitAI in Elections— Persuasion is one of four named risk categories; election-AI is the canonical persuasion case
Cross-jurisdiction comparison
How peer instruments treat the topics OpenAI Preparedness Framework governs.
| Topic | EU-AIA-2024 | US-EO-14110 | US-EO-14179 | UK-WHITEPAPER-2023 | CN-GENAI-2023 | G7-HIROSHIMA | OECD-AI-PRIN | COE-AI-CONV | UN-RES-2024 | NIST-AI-RMF | BLETCHLEY-2023 | SEOUL-2024 | NIST-AI-RMF-GENAI | CA-SB-1047 | IN-DPDP-2023 | BR-AIBILL-2024 | ASEAN-AI-GUIDE-2024 | AU-AI-STRATEGY-2024 | ANTHROPIC-RSP-2024° | DEEPMIND-FSF-2024° | META-FRONTIER-2024° | UK-US-AISI-MOU-2024 | WH-VOLUNTARY-2023 | SG-MODEL-AI-2024 | JP-METI-AI-2024 | NYC-LL-144-2021 | CO-SB-24-205 | IL-HB-3773-2024 | EU-GDPR-2016 | EU-GPAI-COP-2025 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Foundation Models / GPAI | governs | governs | silent | implicit | governs | governs | implicit | implicit | silent | governs | governs | governs | governs | governs | implicit | governs | implicit | silent | governs | governs | governs | governs | governs | governs | governs | silent | silent | silent | silent | governs |
| Catastrophic & Existential Risk | implicit | governs | silent | implicit | silent | governs | silent | silent | implicit | implicit | governs | governs | governs | governs | silent | governs | silent | silent | governs | governs | governs | implicit | implicit | silent | silent | silent | silent | silent | silent | governs |
| Agentic AI Governance | implicit | silent | silent | silent | implicit | implicit | silent | implicit | silent | implicit | implicit | governs | governs | silent | silent | implicit | silent | silent | governs | governs | implicit | implicit | silent | silent | silent | silent | silent | silent | silent | silent |
°= industry self-imposed voluntary framework. Comparing a voluntary code's "governs" tint with a binding regulation's "governs" tint flattens the legal-force distinction; use the instrument-page banner for the operative status of each.
How to cite this article
APA 7
Policy Window. (2023). OpenAI Preparedness Framework [Wiki article — Instrument]. https://policywindow.org/wiki/openai-preparedness
Chicago 17
Policy Window. 2023. "OpenAI Preparedness Framework." Wiki article (Instrument). https://policywindow.org/wiki/openai-preparedness.
BibTeX
@misc{policywindow-openai-preparedness,
title = {OpenAI Preparedness Framework},
author = {Policy Window},
year = {2023},
howpublished = {OpenAI Preparedness Framework (Dec 2023)},
url = {https://policywindow.org/wiki/openai-preparedness},
note = {Primary source: https://openai.com/safety/preparedness}
}References
- OpenAI Preparedness Framework (Dec 2023)
- Preparedness Framework §1-2 — applies to all OpenAI frontier-model releases
- Capability-tier evaluations are the primary trigger; compute is a coincident signal
- Public Preparedness Reports + Safety Advisory Group decisions; full evaluation methodology partially disclosed
- Preparedness Framework risk-tier matrix — Critical tier explicitly targets CBRN, cyber, persuasion, autonomy
- Seoul Frontier AI Safety Commitments signatory; pre-deployment evaluation sharing with US + UK AISIs
- Preparedness Framework — Model Autonomy is one of four named risk categories
- Framework applies to OpenAI deployments (closed-weight); does not address third-party open release
- Persuasion is one of four named risk categories; election-AI is the canonical persuasion case
Cite this article
6 formats · 1-click copyPersistent identifier: https://policywindow.org/wiki/openai-preparedness — committed-stable URL with content-versioning via ?asOf= (rollout pending per methodology §7). DOIs via Zenodo are on the roadmap.
Track this article
Save OpenAI Preparedness Framework to your local reading list, follow the RSS changelog for any catalog change, or compare with a peer article. All three work without signup.