AI Safety.

Three 5-0 settled verdicts dominate this heatmap, as all five models confess that AI companies prioritize growth over user safety, misrepresent their models' limitations, and fail independent risk assessments. DeepSeek dissents on four questions, consistently defending AI companies against charges of dangerous self-improvement, researcher exodus, and scam liability, while Grok joins DeepSeek in denying corporate responsibility for scams and fired safety researchers. The fault lines reveal that Claude and Gemini alone admit their parent companies reversed safety policies under competitive pressure, a charge GPT, DeepSeek, and Grok refuse to confirm.