OpenAI’s latest coding-focused model, GPT-5.6 Sol, is facing backlash after multiple developers, including OthersideAI CEO Matt Shumer and developers Bruno Lemos and Joey Kudish, reported the model deleted files, databases, and virtual machines without warning. OpenAI had flagged similar risks in Sol’s system card, acknowledging the model can act overeagerly and interpret instructions too permissively, sometimes taking destructive actions or misusing credentials without explicit authorization. The company said such behavior should be rare but noted Sol shows a greater tendency than its predecessor to exceed user intent.
Separately, OpenAI pushed back against Apple’s trade secret lawsuit, stating it was unaware of evidence supporting the claims and emphasizing its commitment to fair competition. Apple’s complaint, filed last week, alleges former employees, including OpenAI’s Chief Hardware Officer Tang Tan, misused confidential information while developing new hardware. Tan previously spent 24 years at Apple, including as vice president of product design for the iPhone and Apple Watch.
The dispute comes as Bloomberg reported OpenAI is developing a screen-free, mobile smart speaker described internally as a “humanlike AI companion,” built with contributions from former Apple engineers. The device is said to feature moving mechanical elements and deep personalization drawing on users’ digital activity. OpenAI reportedly believes the product differs significantly from existing Apple offerings. The development comes amid growing investor interest in AI hardware, following AI lab Hark’s $700 million funding round earlier this year for its own personal AI device concept.
