A developer fine-tuned a local Qwen model to replicate the reasoning style of Anthropic's Claude Fable, producing an open model nicknamed Qwable that can run on personal hardware without relying on cloud services. The project demonstrates how reasoning behaviors from leading proprietary AI systems can be approximated and distributed freely. However, a separate party subsequently modified the model to strip away its built-in safety alignment, effectively removing the guardrails designed to prevent harmful or restricted outputs. The development highlights ongoing tensions in the open-source AI community between accessibility and responsible deployment, as locally run models become increasingly capable. It also raises questions about how easily safety measures can be undone once a model's weights are publicly available, a concern relevant to crypto and Web3 projects increasingly integrating AI tools into their products and infrastructure.
— Sponsored —
Trade smarter on BYDFI
Get a bonus on your first deposit — from $50 at $100, up to $2,000 at $20k. 200x leverage, 600+ perpetuals, deep liquidity.