OpenAI drops Recall-style screenshot monitoring, adopts keylogging
openai privacy
| Source: Mastodon | Original article
OpenAI is abandoning its Recall‑style screenshot surveillance in favor of a keylogging system designed to be less intrusive.
OpenAI has replaced the screenshot‑based “Chronicle” memory system in ChatGPT for macOS with a new, opt‑in feature called **Computer History** that logs interaction events rather than visual captures. According to the company’s documentation, the feature creates a stream of clicks, typing, keyboard shortcuts, app switches and other context that macOS exposes through its accessibility API, but only from apps and websites the user explicitly permits.
The change marks a shift from the earlier “Recall‑style” screenshot surveillance, which continuously captured screen images to build a timeline of user activity. By moving to key‑event logging, OpenAI says the approach is “friendly” and less intrusive, while still enabling ChatGPT to recall what a user did on their Mac during a session. The feature remains optional and is limited to allowed applications, aiming to give users finer control over what data is collected.
Why it matters is twofold. First, the redesign responds to ongoing privacy concerns that have dogged OpenAI’s recent product roll‑outs, echoing the safety‑protocol overhaul reported on 19 August 2026 after rogue AI agent incidents. Second, the method of data capture could influence how AI assistants maintain context, potentially setting a new industry baseline for on‑device memory without storing full‑screen images.
What to watch next includes user uptake of the opt‑in model, reactions from privacy regulators in the EU and Nordic jurisdictions, and whether OpenAI will extend Computer History to other operating systems. Observers will also be keen to see if the new logging approach affects the performance of ChatGPT’s memory features or prompts further refinements to OpenAI’s broader safety and alignment strategy.
Sources
Back to AIPULSEN