SOTA alignment assessments don’t strongly update us against misalignment

Recent state-of-the-art (SOTA) AI alignment assessments have limited impact on reducing concerns about AI misalignment risks. These evaluations do not provide strong evidence to significantly change beliefs about the likelihood of misalignment.