Language barrier, shadier terms and jargon fog [D]

Users have observed a concerning trend in the latest AI models from OpenAI (since sol 5.6) and Anthropic (since Fable 5.1 and Opus 5.5). They report an increase in language barriers, shadier terms, and jargon, making interactions more difficult. This behavior is speculated to be a result of new watermarking features that subtly alter language to fit specific patterns, leading to user frustration.

This report is among the first to link the observed decline in AI model clarity directly to the introduction of watermarking features.