From OpenAI concerning their new gpt-oss open-weight language models: ESTIMATING WORST-CASE FRONTIER RISKS OF OPEN-WEIGHT LLMS, in which they tried to make it hack by giving it a terminal in a container and limited web access, but it wasn't very good with computers (compared to o3 and some humans).

gpt-oss blog post: https://openai.com/index/introducing-gpt-oss/
paper blog post: https://openai.com/index/estimating-worst-case-frontier-risks-of-open-weight-llms/
paper: https://cdn.openai.com/pdf/231bf018-659a-494d-976c-2efdfc72b652/oai_gpt-oss_Model_Safety.pdf

It wasn't a good bio-terrorist either, unlike Leah.

#InformationSecurity #InfoSec #CyberSecurity #Hacking #CaptureTheFlag #AI #GenerativeAI #LargeLanguageModels #LLM #OpenAI #GPT #GPTOSS #OpenWeight #MaliciousFineTuning #MFT