Estimating worst case frontier risks of open weight LLMs
OpenAI Blog
August 5, 2025
In this paper, we study the worst-case frontier risks of releasing gpt-oss. We introduce malicious fine-tuning (MFT), where we attempt to elicit maximum capabilities by fine-tuning gpt-oss to be as capable as possible in two domains: biology and cybersecurity.
Verticals
airesearch
Originally published on OpenAI Blog on 8/5/2025