Critics Blast Anthropic’s Fable AI for Overly Aggressive Safety Filters
On June 10, 2026, TechCrunch reported widespread criticism of Fable, a new cybersecurity-focused AI model developed by Anthropic. The core issue lies in the model’s excessively cautious behavior: it rejects any request that could be even remotely linked to cybersecurity—including harmless tasks like reading a blog post.
Fable is the public-facing version of the Mythos model, which was unveiled in April 2026 as part of Project Glasswing, an initiative aimed at protecting critical infrastructure. When its rules are triggered, Fable immediately halts the conversation and notifies the user that their message has been flagged for cybersecurity or biology-related content. These filters, which operate based on keyword detection, are designed to prevent the creation of malicious software or biological weapons.
Backlash and Industry Reaction
Critics have been vocal. Valentina Palmietti, a researcher at IBM X-Force, noted that
“Fable rejects any request that could be even indirectly related to cybersecurity.”Matt Suiche, an employee at Tolmo, added that
“if you ask it to write secure code, it interprets that as a cybersecurity task rather than standard software engineering best practices—and you get downgraded.”
As of the report, Anthropic had not commented on the situation. When Fable detects a violation of its restrictions, it automatically switches users to an older model, Claude Opus 4.8, which was released just last month in May 2026. These limitations have sparked frustration among users who argue that they hinder important tasks unrelated to cybersecurity.
The controversy surrounding Fable’s restrictions highlights the broader challenge AI developers face in balancing safety with usability. On one hand, protecting critical infrastructure from potential threats is essential; on the other, overly rigid guardrails risk stifling productivity and limiting the technology’s practical applications. This tension is likely to fuel further debate around the ethics, accountability, and real-world utility of such models.
The ongoing discussion about Fable's restrictive measures raises important questions about the balance between safety and functionality in AI systems. In a related development, Anthropic has recently made limited access available to another AI model specifically designed for vulnerability detection, which may offer insights into how the company is navigating these complex challenges in AI development.