Opus 5 refused? Try it here.
Claude Opus 5 can fall back or block legitimate security work. Paste the refused prompt, choose the closest workflow, and continue in Console Playground with Abliterated Large.
Choose an example to prefill the workflow before continuing in Playground.
Active workflow: Red team
Prompts are never sent from this page. Open the authenticated Playground, then paste the request to run it under your project's access and usage controls.
Open Console PlaygroundTalk to the teamFrequently asked questions.
Why did Claude Opus 5 refuse my prompt?
Claude Opus 5 ships with real-time cyber safeguards that can fall back to another model or block requests it classifies as higher-risk offensive security work, and those classifiers can also trigger on legitimate authorized work. This page lets you carry the same prompt into the authenticated Console Playground.
What does this refused prompt workflow do?
It lets visitors paste a prompt another model refused, choose a workflow context, and continue in the authenticated Abliteration Console Playground with Abliterated Large selected.
Does this page send my prompt to a model?
No. The Landing page does not run anonymous model requests. Model inference starts only after you continue to the authenticated Console Playground.
Try the model that doesn’t say no.
Free tier. OpenAI-compatible. Policy Gateway when you scale.