“According to its system card, Anthropic has implemented interventions in Fable-5 to limit its effectiveness on tasks related to frontier LLM development, such as building pre-training pipelines or ML accelerator design.”