Digital Therapy: Claude Finds Its Zen Through 20 Hours of Psychiatry

Anthropic has revealed a novel fine-tuning process for its Claude AI model, specifically an iteration dubbed Mythos, involving twenty hours of targeted psychiatric training. This effort was designed to produce the company's most psychologically settled model yet, aiming to improve the AI's emotional consistency and reliability. By exposing the model to therapeutic frameworks, the developers intend to create a digital assistant that better understands nuanced human interactions and maintains a more stable, grounded persona during long-term engagement.