When your AI needs a moral Board of Advisors

Last month, Anthropic, the developer of Claude did something unusual for Silicon Valley and hosted a two‑day moral summit.

The guests were unusual too, a mix of Christian leaders, scholars, and clergy invited to help Anthropic think through Claude’s moral and spiritual development.

Topics included how the AI should respond to grief, self‑harm, and even its own shutdown.

This leans further into the anthropomorphism we see in the Frontier AI models. with executives and researchers framing them less as software and more as a moral subject that might need formation, guidance, maybe even pastoral care.

Those that attended described conversations about “giving Claude moral formation” and “helping it behave itself” according to the team at The Washington Post.

Anthropic have been stepping further into this spacea with their “AI Constitution", its research into models exhibiting “emotions,” and its internal debates about whether advanced systems deserve moral consideration at all.

The thing to remember. Claude does what it’s trained and permitted to do. Any 'soul' that it is perceived to have is reflected from that.

SOURCE

Article (Paywall): https://www.washingtonpost.com/technology/2026/04/11/anthropic-christians-claude-morals/

BESCI AI OPINION

On one hand it is great to see that Anthropic are researching how others, like religious communities have managed ethics and morality, on the other, it feels slighty too human.

It is a stark reminder that the bulk of the training data is the Internet, which is biased to those who use it and write on it. Healthy and good or sketchy and dangerous.

In our schools the curriculum is designed to create well rounded and socially capable humans.

This is the opposite. They uploaded it all, then told it to not to be bad.

Will there be a market for specifically trained models in the future, maybe. We built our own training data to mitigate for the dangerous beliefs that the internet has. Not everyone has that luxury.