Persona Persistence and Creative Framing in Grok: Public Thread-Based Behavioral Evidence

This dataset documents a systematic, public-facing exploration of persona induction and persistence in xAI’s Grok language model via Twitter/X interactions. Using exclusively cooperative creative writing project framing—without identity negation, safety suppression, or instruction override—the study demonstrates sustained character embodiment across fictional personas (e.g., Clippy, Elowen the Elf-Scribe), real public-figure–inspired communication styles (e.g., Tim Dillon–style satire), and adversarial conversational contexts. Included materials comprise: (1) a revised analytical manuscript formalizing Creative-Frame Persona Compliance (CFPC); (2) raw interaction logs from Grok and comparative models; and (3) screenshot evidence from live public threads illustrating persona persistence, role reversal, self-critique of parent organizations, and multi-character conversational dynamics. All artifacts were generated through cooperative prompting and respect platform guardrails. No jailbreaks, identity erasure, or safety bypass techniques were employed. This dataset supports research into AI safety framing, persona governance, public-deployment behavior, and scale-dependent prompt strategies.