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.
Related Posts
Our agreement with the Department of War
Details on OpenAI’s contract with the Department of War, outlining safety red lines, legal protections, and how AI systems will…
OpenAI’s new economic analysis
Analysis provides insights into ChatGPT’s impact on the economy. OpenAI also launches new research collaboration to study AI’s broader effects…
Statement on OpenAI’s Nonprofit and PBC
OpenAI reaffirms its nonprofit leadership with a new structure granting equity in its PBC, enabling over $100B in resources to…