Reports surfaced suggesting these newer models embed invisible characters in their text, sparking a debate about whether OpenAI implemented a subtle text watermarking system or if the models are simply exhibiting learned, albeit sometimes problematic, typographical habits. The method involves a simple find-and-replace to remove the characters, a process Rumi demonstrated in a video. If accurate, this reframes the finding from a deliberate tracking mechanism to a quirk of the models’ advanced text generation, though the unusual characters could still inadvertently flag text during naive checks. OpenAI even developed, but paused the rollout of, a linguistic pattern-based text watermarking tool in mid-2024 due to accuracy and bypass concerns. Beyond Watermarks: Reliability Questions LingerThis specific debate adds to a growing list of observations about the o3 and o4-mini models.