Anthropic ends hidden Fable throttling; OpenAI weighs price cuts to block Anthropic discount war
Anthropic replaced covert response degradation on frontier-model queries with explicit API refusals and Opus 4.8 fallback. OpenAI is considering sharp price cuts to preempt Anthropic's own discount strategy and lock in enterprise revenue.
Anthropic reversed a hidden policy that quietly degraded Fable responses when users asked about developing competing large language models or frontier AI systems. The API now returns an explicit refusal—matching the treatment of other filtered topics—and falls back to Opus 4.8, with a specific reason code included in the response. The shift addresses user complaints that Fable was silently underperforming on technical queries about model architecture and training pipelines without any transparency.
OpenAI is weighing aggressive price cuts to preempt what intelligence suggests is Anthropic's own discount strategy. The concern inside OpenAI is that Anthropic could move first and capture enterprise customers comparing vendors on cost-per-token. If Anthropic hesitates, OpenAI sees an opening to lock in revenue share before a full discount war begins. Both companies have held list prices stable since late 2025, though enterprise contract negotiations have already pushed per-seat discounts above 30 percent in competitive deals.
