§ feed · storyline
Estimating worst case frontier risks of open weight LLMs
OpenAI publishes research estimating worst-case frontier risks of open-weight LLMs by testing malicious fine-tuning in biology and cybersecurity domains.
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.
§ sources1 publication · timeline below