On January 16, OpenAI announced ChatGPT would begin showing advertisements, with ads live by February 9. Just eight months earlier, the company spent $6.5 billion acquiring Jony Ive's hardware startup io to build a screenless, always-sensing device designed to replace smartphones. But OpenAI is just the latest entrant in a troubling trend.

Every company building AI assistants is now funded by advertising, creating a structural conflict as these devices become increasingly intimate. The industry is moving toward always-on AI that continuously listens and watches, abandoning the wake-word model that dominated since 2014. Real conversations—like parents discussing groceries and school schedules in a busy kitchen—happen without trigger phrases, making continuous sensing essential for truly helpful AI assistance.

Local on-device inference offers the only architectural solution to this privacy crisis. Modern hardware can run complete AI pipelines locally without transmitting sensitive data, using open-source models and efficient compression. The question isn't whether always-on AI will happen, but who controls the data. With advertising companies controlling the infrastructure, the intimate details of our lives become commodities. The only safe architecture is one where devices physically cannot transmit the data they process, regardless of policy promises.