Anthropic has implemented stricter safeguards on its newly launched AI model, Claude Mythos 5, leading to significant backlash from advanced users who find these restrictions overly limiting. Users have reported frustrations, noting that even mentioning benign terms like “cancer” can trigger the model’s safety protocols, hindering their research efforts. These enhancements were introduced when Anthropic launched Claude Fable 5 and Mythos 5 in June 2026, aiming to balance powerful capabilities with safety, especially in high-risk areas, although this approach has resulted in reduced utility for researchers.

Anthropic: Anthropic is an AI research company focused on developing advanced large language models, primarily the Claude family. In early June 2026, it released Claude Fable 5 as its most capable publicly available model, paired with stricter safety classifiers that route sensitive queries on topics like biology, chemistry, and cybersecurity to less powerful fallback models. This conservative approach to safeguards, tuned to enable a quicker safe release, has drawn immediate criticism from advanced users for limiting legitimate queries.

Model Launch: Anthropic launched Claude Fable 5 and the more restricted Claude Mythos 5 in June 2026, making powerful Mythos-class capabilities available to the public for the first time while reserving broader access for vetted partners via Project Glasswing.
User Reaction: Advanced users and AI researchers have expressed frustration with the restrictions, reporting that even innocuous terms or tasks trigger fallbacks and reduce the model’s utility for research and development work.
Safety Approach: The company has tuned its new safeguards conservatively to prioritize safety in high-risk domains, acknowledging that this can sometimes intercept benign requests as the systems are refined.