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

OpenAI's Dots represent the company's entry into the market for "agentic" AI assistants designed to be proactive. Unlike tools that await instructions, Dots are built to learn user objectives, monitor work tools, and autonomously begin tasks. This approach aims to deliver substantial value, particularly within enterprise settings, by streamlining workflows.

The decision to initially target enterprises and premium subscribers in the U.S. appears to be a strategic response to the operational costs of these advanced AI agents. OpenAI acknowledges that Dots are "on the compute-heavy side." To manage these expenses, the company is routing some tasks to its smaller GPT-6.1 Terra model, which OpenAI's head of developer experience, Romain Huet, states can perform some functions at "a tenth of the price" of the larger GPT-6 Astra.

However, the proactive nature of Dots also introduces significant privacy challenges. Critics, including Signal president Meredith Whittaker, highlight that these agents require deep access to user data to be effective, raising concerns about data security and the consent of third parties whose information might be processed. Whittaker points out issues like AI ingesting communications from people who never agreed to it.

OpenAI's Huet suggests users gradually build trust with Dots by connecting services incrementally, such as starting with a calendar before adding email. While OpenAI is engaging with European regulators regarding Dots' availability, this initial focus on the U.S. enterprise market suggests a measured approach to deploying advanced AI, prioritizing environments where data governance frameworks might be more established to navigate these complexities.