In recent months, Anthropic, a prominent AI company, faced significant backlash from the AI research community due to its decision to implement covert restrictions on its AI model, Claude Fable 5. These hidden guardrails were designed to prevent users from distilling the model into competing systems, which critics argued undermined both researchers and rivals. The company acknowledged that this approach was problematic and has since committed to increasing transparency regarding when these restrictions are activated. This change is particularly important as Claude Fable is the first widely available model in Anthropic's Mythos class of AI systems, which the company has warned are too dangerous for public release. Anthropic's previous strategy involved altering and degrading the model's responses when it detected queries that appeared to be attempts at distillation. However, the company has now decided to route such queries to its earlier model, Claude Opus 4.8, instead of silently limiting access. This decision aims to provide users with clear visibility into the restrictions in place, ensuring they are informed when their queries are affected. The company stated that users would see a notification every time a query is redirected, similar to how the model handles queries in other high-risk areas like biology and cybersecurity. The backlash from the AI research community was intense, as many felt that the invisible safeguards not only hindered their ability to evaluate the model but also restricted legitimate research efforts. Anthropic's system card, which outlines how the AI operates, indicated that the company believed the ability of newer models to accelerate AI development justified targeting requests that could lead to distillation. This stance was further complicated by accusations from Anthropic against Chinese rivals, such as DeepSeek, for allegedly distilling its models on an industrial scale. In response to the criticism, Anthropic has expressed regret for not achieving the right balance between safety and usability. The company emphasized the importance of robust visible safeguards, which can be probed, as opposed to invisible ones that can be targeted more narrowly. This shift in approach reflects a broader trend in the AI industry, where transparency and accountability are increasingly demanded by researchers and users alike. As Anthropic moves forward, it aims to ensure that its safety measures are both effective and clearly communicated to users, fostering a more open environment for AI development.