Anthropic's Fable 5 Sparks Outrage Over Excessively Strict Safety Barriers

Artificial intelligence (AI) giant Anthropic is facing a wave of sharp criticism following the launch of its latest model, Fable 5. Although touted as its most powerful public release to date, the extremely strict safety system has triggered a negative reaction from users and AI researchers. Fable 5, part of the Mythos model family, debuted on Tuesday (9 June 2026). However, user excitement quickly turned to disappointment. Many parties believe Anthropic has applied excessive guardrails, degrading the quality of answers without informing users. Anthropic argued that strict restrictions are necessary to prevent misuse of the highly powerful model. The company fears a model like Mythos could be used by malicious actors to launch cyberattacks or design biological weapons. As a preventative measure, Fable 5 was programmed to refuse questions related to cybersecurity and biology. However, initial testing showed the system often misidentified harmless requests. Fable 5 reportedly refused to answer general medical questions, such as examinations for pancreatic injuries, and declined to offer opinions on public figures like Elon Musk and Anthropic’s own CEO, Dario Amodei, citing safety reasons. The sharpest controversy arose over Anthropic’s policy of deliberately lowering the intelligence of answers if the query related to AI development. This was done to prevent competitors from using Anthropic’s technology to accelerate their own research. Initially, this restriction was invisible. The policy was condemned by experts. Clement Delangue, a prominent AI researcher, called the action the highest form of manipulation. Responding to public pressure, Anthropic ultimately backtracked less than two days after the launch. ‘You should have visibility into the protections we apply and why. We apologise for not getting the balance right,’ Anthropic wrote via their official account on X on Thursday morning. Peter Wallich of the Constellation Institute assessed that while the launch was problematic, Anthropic’s approach was reasonable given the power of the technology they possess. According to him, releasing a model with overly loose protections would be far more irresponsible and could potentially cause irreparable harm. The Fable 5 case underscores the challenges technology companies face in determining the boundary between public safety and users’ rights to transparently access beneficial information.