Deliberative alignment: reasoning enables safer language models

OpenAI introduces a new alignment strategy called deliberative alignment for o1 models, which teaches safety specifications and reasoning over them.

From the source
Introducing our new alignment strategy for o1 models, which are directly taught safety specifications and how to reason over them.
openai.com

Visit source ↗openai.com

Announced: 20 Dec 2024 10:00 UTC
Detected: 03 Aug 2026 21:11 UTC (+591d after)

Other stories

OpenAI · 31 AugPolimill builds Japan's next-generation public AI infrastructure

OpenAI · 31 AugA milestone in expanding access to AICompany news

Anthropic · 31 AugImproving our alignment and security effortsSafety or policy

OpenAI · 28 AugOur decision on Cursor following its acquisition by SpaceX

OpenAI · 28 AugSupporting Thailand’s next generation of AI startups

Partnership

OpenAI · 27 AugBetter answers, broader thinking: What students gain from ChatGPT and critical-thinking trainingResearch paper