Security scan: pick any model, or use your own key
The scan dialog now recommends the newest strong models and lets you search every model your account can use, including models on your own API key.
Anyone who runs security scans can now pick any text model, including recommended current ones or a model on their own API key. No action needed.
Text version
Recommended: Kimi K3.1 Pro as the default, DeepSeek V4.1 Flash 1017, GLM 5.3, an uncensored Qwen 3.8 27B, Kimi K3 and DeepSeek V4 Pro. All models: a search over every model your account can use in chat. Your own key: models from providers you connected with your own API key.
The scan dialog used to offer three models. It now opens with a recommended list of six: Kimi K3.1 Pro as the default, DeepSeek V4.1 Flash 1017, GLM 5.3, an uncensored Qwen 3.8 27B, and the earlier Kimi K3 and DeepSeek V4 Pro. The uncensored model is there on purpose. A security scan asks for attack payloads, and a model that refuses half of them finds less.
Below the list you can search every model your account can use in chat. That includes models from providers you connected with your own API key, such as Anthropic, OpenAI or DeepSeek. Those run on your key, and ZeroLog bills no AI cost for them. The scan fee stays the same.
A model outside the recommended list has not been tested for scans. A model that answers poorly structured stops earlier, and the dialog says so. If you pick a model your account cannot use, the scan is refused before you pay, with the reason.
No action is needed. Scans you start without changing the model run on Kimi K3.1 Pro.