Anthropic, led by Dario Amodei, struggles to balance AI safety advocacy with developing powerful, risky models like 'Mythos.' Despite engaging global institutions to manage impact, developers voice unease about increasing AI autonomy and opacity, raising accountability concerns. The article highlights Anthropic's "walking contradiction," advancing AI rapidly while publicly positioning itself as responsible. This difficult "juggling act" of trust and innovation intensifies as its tools become more independent, challenging principles and market ambitions.