Anthropic Responds to Complaints of Claude Getting 'Dumber' — Says the Model Isn't at Fault

A growing wave of user complaints has swept through Claude's community in recent days, with many users reporting that the AI assistant's responses have become noticeably less sharp, more conservative, and more template-driven. Social media and online forums have lit up with comparisons, benchmarks, and mounting frustration over what users describe as Claude getting "dumber."

Anthropic has now formally responded to these concerns, stating clearly that the underlying Claude model has not undergone performance degradation. According to the company's response covered by Chinese tech media 36Kr, the issue is not about the model's intelligence declining.

Instead, Anthropic suggested the perceived quality drop likely stems from non-model changes — such as updated system prompts, adjusted safety filter sensitivity, or modifications to context processing mechanisms. These system-level adjustments may have altered Claude's output style and reasoning depth in certain scenarios, creating the impression that the model itself has become less capable.

This is not the first time Anthropic has faced user pushback over perceived quality changes. As a company built around AI safety, Anthropic takes a cautious approach to model behavior controls, and safety-related system adjustments can sometimes unintentionally impact response openness and depth. Similar debates erupted after the Claude 3 family launch when users noticed stylistic shifts.

The core tension here is familiar across the AI industry: when safety considerations and user experience come into conflict, where is the right balance? Anthropic's safety-first posture is well-documented, but the user backlash demonstrates that any system-level fine-tuning risks being interpreted as a degradation of the model itself.

It remains unclear whether Anthropic plans further adjustments to address user experience concerns. The situation bears watching, particularly to see if Anthropic discloses more details about recent system changes — and whether competitors like OpenAI and Google face similar user reactions to their own safety adjustments.