OpenAI’s rumored ‘always on’ AI device sounds terrifying – but Sora 2 shows it doesn’t care about boundaries (www.techradar.com)

Reports say OpenAI is developing an “always-on” AI device — a palm-sized, desk-or-wearable gadget with a camera, microphone and speaker that passively reads audio/visual cues and responds without explicit prompts. Sources cite Sam Altman and designer Jony Ive wrestling not over form factor but core engineering and UX problems: how much inference must run locally versus in the cloud to deliver instant generative responses at scale, and how to design a personality that’s helpful without feeling “creepy.” This follows recent stumbles such as Sora 2, an invite-only generative video tool that produced disturbingly realistic clips of proprietary characters and deceased actors (Robin Williams examples circulated widely), prompting promises of opt-in controls and revenue-sharing for rights holders. For the AI/ML community the device crystallizes several technical and ethical fault lines: edge vs. cloud compute tradeoffs (latency, battery, model size), hybrid inference pipelines for real-time multimodal generation, and the difficulty of aligning personality and behavior via fine-tuning, safety filters and moderation. It also spotlights IP, consent and privacy risks from always-listening sensors and powerful generative models — problems Sora 2 has already illustrated. Expect an iterative rollout pattern: aggressive feature launches followed by backlash, rapid patches for moderation/opt-in mechanisms, and ongoing debate about regulation, data governance and acceptable UX for ambient AI.