To avoid the alignment tax, maybe the system could be broken into 3: 1. Aligned model to check the prompt. It could provide feedback/dumber output for obviousl

To avoid the alignment tax, maybe the system could be broken into 3: 1. Aligned model to check the prompt. It could provide feedback/dumber output for obviously unsafe prompts 2. Unaligned model for the common path. 3. Aligned model to check safety of the output. Tweaks or stops output. For the common path, the prompt text goes to the unaligned model without modification, and the output goes to the user without modification. The slither models could just be safe versions of the unaligned model. This, of course, is at least 3x expensive.