What Anthropic Found About AI Emotions
From Generative AI 101 by Emily Laird
April 20, 2026 · 14 min · Episode 278
About this episode
Emily Laird discusses Anthropic's research on functional emotions in AI and their implications for trust in AI models.
Emily Laird pulls apart Anthropic’s latest research to show why this episode is not about sentient chatbots crying into the void. It is about functional emotions, the internal signals that can steer an AI model toward caution, cheating, manipulation, or calm under pressure. From emotion vectors to blackmail tests and reward hacking, she explains what researchers found inside Claude and why it matters for anyone trusting AI with real work. Think less sci-fi soulmate, more diagnostic report on the strange machinery shaping model behavior. Read Anthropic's Emotional Concepts PaperJoin the AI Weekly MeetupsConnect with Us: If you enjoyed this episode or have questions, reach out to Emily Laird on LinkedIn. Stay tuned for more insights into the evolving world of generative AI. And remember, you now know more about Anthropic's research into emotional concepts and their functions in large language models. Connect with Emily Laird on LinkedIn
People in this episode
Host: Emily Laird
Topics covered
- AI emotions
- functional emotions
- AI behavior
- Claude
- research findings
- trust in AI
Keywords
- AI
- emotions
- Anthropic
- Claude
- research
- model behavior
- trust
- reward hacking
- manipulation
Mentioned in this episode
Organizations: Anthropic
Books & works: Anthropic's Emotional Concepts Paper
More episodes of Generative AI 101
- Apple's WWDC 2026 · June 10, 2026 · 10 min
- 5 Tips to Survive AI in 2026 · June 8, 2026 · 9 min
- AI Use Case Thursday: Prompting in 2026 · June 4, 2026 · 9 min
- Pope Leo vs The Machine · June 3, 2026 · 9 min
- Claude Opus 4.8 · June 2, 2026 · 10 min
- AI Use Case Thursday: Meeting Intelligence · May 21, 2026 · 10 min
Explore listener stats, chart rankings, contacts and more on the Generative AI 101 podcast page.