We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
OpenAI Blog·Dec 11·3 min read
Update to GPT-5 System Card: GPT-5.2
TL;DR
OpenAI released GPT-5.2, an updated model in the GPT-5 series, to address the need for transparent documentation of safety practices in large language models. The update maintains the same safety mitigation approach (risk reduction strategies) described in prior GPT-5 system documentation, trained on diverse public internet data, third-party partnerships, and user-generated content.
✦ Why It Matters
Engineers can reference GPT-5.2 system documentation to understand current safety practices and compliance standards for production LLM deployments.
Key Takeaways
- GPT-5.2 maintains safety practices consistent with GPT-5 and GPT-5.1 documentation.
- Training data combines public internet sources, licensed third-party partnerships, and human-generated content.
- Review system card documentation to understand safety evaluation and risk mitigation strategies.
- This is a documentation update confirming consistency, not a feature release with new capabilities.
Full Summary
GPT-5.2 represents an incremental update to OpenAI's GPT-5 model family, focusing on maintaining transparency around safety practices rather than introducing novel capabilities. Large language models (LLMs)—AI systems trained on vast text datasets to generate human-like responses—require documented safety mitigations (techniques to reduce harmful outputs and misuse risks).
GPT-5.2 was trained using a multi-source approach: publicly available internet data, licensed third-party datasets, and content generated or provided by human trainers and researchers. The safety mitigation strategy remains aligned with frameworks outlined in the GPT-5 and GPT-5.1 system cards, which detail how the model was evaluated and safeguarded against potential harms.
This update emphasizes continuity in safety practices across model iterations rather than announcing breakthrough improvements. For practitioners, this signals that safety documentation standards are being maintained as models evolve.