Under the guise of preventing Altman from becoming corrupt, former employee reveals that Anthropic has fallen deeply into the myth of being a 'good person' According to Beating monitoring, Anthropic, which claims to have "high trust, low self", is sliding towards ideological homogenization. The report by WIRED reveals internal disagreements within the company regarding governance evaluation. A former employee revealed that there is an active debate atmosphere within the company; But other former employees have depicted a more dull scene, stating that sharp criticism is largely isolated in private group chats, and regular staff meetings are even jokingly referred to as "Dario Vision Quests" by employees. Listening to CEO Dario Amodei's speech is like listening to a sermon, and few people challenge management decisions in public meetings. The root of the dull atmosphere lies in Anthropic's elite security belief of "using attack to promote defense". Internally, executives vigorously view OpenAI CEO Sam Altman (as well as Meta and xAI) as a negative example, while instilling elite logic in employees that they must maintain industry leadership in business, computing power, and research talent in order to be qualified to have a say in rule making. Former OpenAI director Helen Toner once made a forest analogy: Villagers will flock into a magical forest filled with monsters no matter what, while Anthropic's strategy is to go deep into the forest first, release technological dividends while taming monsters and controlling catastrophic risks. The pursuit of technological dominance has also led to controversy over Anthropic's safety and ethical choices. In the autumn of 2024, despite internal opposition, the management teamed up with Palantir to open up services to US intelligence and defense agencies. At present, Claude has been confirmed to be used for target recognition in the Middle East conflict. In the face of questioning about attacks that have caused civilian casualties, Amodei, although not aware, argued that "as long as humanity makes the final decision, the use of technology is compliant". In June of this year, the development team secretly embedded destructive code in the flagship model Claude Fable 5 to secretly interfere with illegal development. But after facing strong protests from the industry, the company was forced to compromise and announced that it would adjust its secret interference mechanism to publicly visible security restrictions. Industry criticism points out that completely betting on the moral superiority of a few elites for cutting-edge AI security will inevitably lead to unavoidable blind spots in self-discipline. [Original link]