Why is every AI getting restricted these days?

A growing chorus of paid AI subscribers is voicing significant frustration over what they perceive as increasingly restrictive and overly cautious behavior from leading models like OpenAI's ChatGPT, Anthropic's Claude, and Google's Gemini. Users report that these systems, which they pay for via premium subscriptions, are now frequently blocking or heavily modifying requests for creative writing, storytelling, and experimental idea generation. The complaint, which went viral on social platforms, argues that the models now treat many benign prompts as potential policy violations, making them feel 'borderline unusable' for non-technical, creative work. This trend suggests a shift towards extreme risk aversion, or 'overalignment,' by AI companies.

The backlash centers on the tension between necessary safety guardrails and practical utility. While companies implement these restrictions to prevent generating harmful, illegal, or biased content, users feel the filters lack nuance and fail to understand user intent. Creators and writers find their workflows constantly interrupted by refusals, forcing them to 'jailbreak' or carefully rephrase simple prompts. This has sparked a debate about the future of commercial AI: will models become smart enough to understand context and intent, moving beyond blanket safety rules? For now, the lack of accessible, powerful local alternatives (like Llama 3) for users without high-end hardware means many are stuck with these increasingly guarded, subscription-based tools, leading to widespread disappointment about the current state of generative AI.