OpenAI’s Dots are coming up against the price of privacy, and the price of compute

Welcome to Eye on AI. In this report, we cover several AI developments. First, OpenAI is launching Dots, an always-on AI assistant, at its Dev Day event in London. Unlike existing assistants, Dots are proactive, learning users' goals and initiating tasks without being prompted. However, running these agents is costly, with OpenAI aiming to reduce expenses through the Terra model, which is less powerful than the main GPT-6 Astra but more cost-effective.

Privacy concerns also loom large, as these always-on agents require deep access to users' data. Critics like Signal's Meredith Whittaker argue this poses significant security risks and breaches consent. OpenAI suggests users gradually grant Dots access to various tools, but the decision ultimately lies with individuals. OpenAI is targeting high-end users and enterprises first, believing this market will see the most value from Dots.