Open-weight models
Also called: open-source models · open models
Open-weight models are AI models whose trained weights are published for anyone to download, run and fine-tune, such as Llama, Qwen, Mistral, DeepSeek and GLM releases. Their safety controls can be removed by fine-tuning, so their capabilities, especially in cyber offense, matter for security.
How it works
Once weights are public, refusals and safeguards can be trained away and the model run without monitoring. Security work on open weights looks at what the models can do unaided, how easily safeguards are stripped, and whether downloaded weights are what they claim to be.
How to defend
Assess open models' capabilities before deployment, verify weights and publishers, and do not assume vendor safety tuning will hold once a model is in someone else's hands.
Real cases 2 stories
INFO
Research · arXiv
SLDR: restoring refusals after a model is fine-tuned for harmSLDR is a defense against malicious fine-tuning that trains a LoRA recovery adapter on the safety-sensitive layers only and switches it on for harmful queries. The authors report large cuts in harmful output across four model families.
INFO
Benchmark · NIST CAISI
CAISI: GLM-5.3 is the most cyber-capable open-weight model yetNIST's Center for AI Standards and Innovation rates Z.ai's GLM-5.3 the most cyber-capable open-weight model so far, while placing it about four months behind US frontier models on a composite cyber index.