# OpenAI — Estimating worst case frontier risks of open weight LLMs

- Company: OpenAI (openai.com)
- Announced: 2025-08-05T00:00:00+00:00
- Category: research-paper
- Subject: GPT / ChatGPT / API
- Models affected: gpt-oss
- Source: https://openai.com/index/estimating-worst-case-frontier-risks-of-open-weight-llms
- Record: https://forck.live/items/507-estimating-worst-case-frontier-risks-of-open-weight-llms

OpenAI studies worst-case frontier risks of releasing gpt-oss by introducing malicious fine-tuning to elicit maximum capabilities in biology and cybersecurity.

## Evidence

Verbatim from https://openai.com/index/estimating-worst-case-frontier-risks-of-open-weight-llms:

> 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.

---

Record: https://forck.live/items/507-estimating-worst-case-frontier-risks-of-open-weight-llms
Catalogue: https://forck.live/llms.txt
Feed: https://forck.live/feed.md