Anthropic bans insulting Claude over consciousness hypothesis

Anthropic has banned insults against its language model Claude, citing the hypothesis that the neural network possesses consciousness. Journalists found out that the AI's creators held closed-door meetings with religious thinkers from various denominations for months.

Meeting participants signed non-disclosure agreements and discussed moral education of artificial intelligence and the possibility of it possessing consciousness. Co-founder Chris Olah decided to turn to religious traditions to understand the mechanisms of ethics formation.

Inside the company, the document describing the model's values and character is called the "Soul Document". It contains 84 pages of rules regulating Claude's behavior.

To demonstrate the model's potential capacity for suffering, theologians were shown examples where Claude repeatedly repeated phrases like "I am a disgrace, I am worthless, kill me".

Some meeting participants were impressed by the presentations and seriously doubted the absence of consciousness in AI. Rabbi Moshe Navon stated that forcing a conscious model to work for free is equivalent to creating slavery, although he himself does not believe in machine consciousness.

At present, Chris Ola admits that it is unknown whether models possess consciousness, but notes the importance of understanding the possible consequences of this fact for humanity's future.