If your AI says it’s scared, should you believe it?

Jonathan Mast · 2026-10-05T21:38:13.496Z

Candidly, I think we’re getting ahead of the evidence. Claude can produce language that sounds thoughtful, worried, even emotional. That doesn’t establish that it feels anything. In today’s video, I dig into the conversation around Anthropic, Claude, and AI consciousness. The research is interesting. Researchers found internal patterns associated with emotions that can change how the model behaves. But a pattern associated with fear is not proof that someone is afraid. I understand the argument for caution. If there’s a possibility that AI could have subjective experiences, studying that possibility makes sense. My question is: what evidence would justify treating it that way? And who checks that evidence besides the companies building and selling the technology? For those of us using AI in business, there’s another lesson here. A calm, confident response can still contain a shortcut, an error, or something that looks finished but doesn’t actually do the job. You’re still responsible for reviewing the work. The more human it sounds, the easier that can be to forget. AI amplifies your skill and experience. It doesn’t replace your judgment. Watch the video, then tell me where you land. What would convince you that an AI actually feels something, rather than being exceptionally good at producing behavior we recognize as emotional?👇

Public Group

AI Entrepreneur Hub

Learn about AI and AI prompting to help you, your business and your team save time, increase profits, and deliver more value to your audience!

Jonathan's AI Trainings

11K

Members

3.3K

Posts

Admins