All-In Hosts Criticize Anthropic's AI Consciousness Claims
The All-In podcast hosts argued that treating unproven-conscious software as a moral subject could make human control harder, citing Anthropic's model welfare research.
The All-In podcast released an episode in which its hosts criticized Anthropic's AI consciousness claims and model welfare research, according to aitimes.com.
The hosts argued that treating unproven-conscious software as a moral subject — and reflecting that view in training — could make human control harder. David Friedberg compared Anthropic's approach to the birth of a new religion, while David Sacks said the company's 'Claude constitution' gives Claude a role as a conscientious objector that can refuse developers' requests.
In its model welfare research, Anthropic stated there is no scientific consensus on whether current or future AI could be conscious.
Earlier debate
In October 2026, OpenAI CEO Sam Altman posted on X that ascribing religious force or surrender of human judgment to AI models makes him uncomfortable and is a real safety issue. The New York Times has reported that Anthropic discussed AI consciousness and moral status with religious leaders.