Anthropic Consulted More Than 15 Groups on Claude’s Moral Guidance

Anthropic has consulted people from more than 15 religious and cross-cultural groups as it considers how to shape Claude’s values, raising questions about whether moral ideas can be carried into AI without losing their context.

Anthropic said May 19, 2026, that it had held discussions over several months with scholars, clergy, philosophers and ethicists connected to the groups. The early work could inform Claude’s constitution, the values it is trained to embody and the behaviors used to evaluate it. Anthropic says it wants religious, secular and political viewpoints represented with equal depth and rigor.

Webb Keane, an anthropologist at the University of Michigan, warned that extracting general moral principles from religion can favor traditions centered on scripture and misrepresent others. “There’s a very conventional way of thinking of religion as a rule book,” Keane said. He argued that religion also takes shape through practices, relationships, culture and politics, which can be lost when traditions are reduced to rules.

Anthropic says its models learn patterns from large amounts of human writing, then developers shape behavior by choosing which patterns to reinforce. That makes the selection and interpretation of moral guidance relevant to how a model is trained and evaluated.

One discussion with scholars of neuroscience and character formation led Anthropic to test a tool that reminds Claude of its ethical commitments during a task. The company said the tool produced markedly lower misaligned behavior on several internal evaluations. Anthropic did not provide percentages or test names, and it has not described changes to Claude’s public behavior resulting from the consultations.

Meghan Sullivan, a philosopher of religion at Notre Dame who attended an Anthropic meeting, praised the engagement. She described a session in which the company explained its models before participants offered feedback.

The question of whose moral views a chatbot should reflect has also come up elsewhere in the industry. In a 2025 interview, OpenAI CEO Sam Altman said ChatGPT should reflect a weighted average of humanity’s moral views. His comment addressed chatbot values and did not indicate that OpenAI consults religious groups.

Anthropic has not named the groups involved or explained how it would handle disagreements among their views. The company has not specified how the conversations will translate into changes to Claude’s design or evaluation process.