- Home
- Encyclopedia
- Frontier safety framework
RSI Encyclopedia · Term
Frontier safety framework
A frontier safety framework is a lab's published plan for spotting dangerous capabilities in its SI models and the safeguards it will apply when they appear.
- Category
- Term
- Also called
- safety framework, Frontier Safety Framework
- Rogue Log cases
- 0
- Laws that use it
- 13
In plain English
At the 2024 Seoul summit most leading labs agreed to publish safety frameworks for their frontier models, and by 2026 almost every major developer had one under some name: Anthropic's Responsible Scaling Policy, OpenAI's Preparedness Framework, Google DeepMind's Frontier Safety Framework, Meta's Superintelligence Scaling Framework, xAI's Frontier Artificial Intelligence Framework, Microsoft's Frontier Governance Framework and Amazon's Frontier Model Safety Framework.
The frameworks share a structure. They list severe risk areas such as chemical and biological weapons, cyber attacks, autonomous replication and automated SI research, define capability levels that count as dangerous, and promise safeguards for each level. They differ on how measurable the thresholds are, whether commitments are binding or discretionary, and who checks compliance. Laws are starting to require them: New York's RAISE Act, for example, obliges large frontier developers to publish such a framework and review it every year.
Why it matters for rogue SI
These frameworks are the labs' own rules for avoiding rogue SI. The safety framework grade on the Containment Board measures how specific and enforceable each one is.
Laws and rules that use it
- In force EO 14365 United States
- Proposed FRONTIER Act United States
- In force SB 53 California
- In force EO N-9-26 California
- Passed, not yet in force RAISE Act New York
- Passed, not yet in force SB 315 Illinois
- Voluntary or guidance Framework 3.0 China
- Voluntary or guidance Voluntary Code Canada
Lab policies
- Responsible Scaling Policy Anthropic · version 3.4
- Preparedness Framework OpenAI · version 2
- Frontier Governance Framework OpenAI · version May 2026
- Frontier Safety Framework Google DeepMind · version 3.1
- Superintelligence Scaling Framework Meta · version 2.1
- Frontier Artificial Intelligence Framework xAI · version June 30, 2026
- Frontier Governance Framework Microsoft · version February 2026
- Frontier Model Safety Framework Amazon · version September 2026
Sources
- Ensuring a National Policy Framework for Artificial Intelligence Federal Register ·
- H.R. 9925, FRONTIER Act, 119th Congress LegiScan
The RSI Encyclopedia
Keep reading
207 entries on the terms, labs, models, laws and events of rogue SI, plus every case in the Rogue Incident Database.
Enter RSI
Every term, lab, model and law
Browse the full A to Z index, filter by category or search for a name.