So the endpoint of 'effective altruism' is the Palantir contract, and internal dissent can only be expressed in small groups. This governance structure is said to be to prevent an OpenAI-style crisis, but it ended up becoming a more covert version.

According to former employee leaks, there are internal governance disagreements at Anthropic, where challenging senior management in public meetings is difficult, and the CEO's speeches have been likened to priestly sermons, with internal debates mostly taking place in private group chats. Management uses OpenAI's Altman as a cautionary example, emphasizing the need to stay ahead in business, computing power, and talent. In the fall of 2024, management partnered with Palantir to open services to U.S. intelligence and defense agencies, and Claude has been used for target identification in Middle East conflicts.