OpenAI Dots Turns Codex Into a $100 Always On AI Assistant
OpenAI's dots turns Codex technology into a $100-a-month proactive assistant, raising privacy, permission, timing and payment access concerns for users.
Summary
OpenAI released dots at its late September 2026 developer day, joining Instinct and Meta’s heavily marketed Muse in the race for persistent AI assistants. Available through ChatGPT’s desktop and mobile apps to Pro subscribers or higher, from $100 monthly, dots connects to calendars, email, computers and other tools, then acts without prompts. Enterprise product leader Alexander Embiricos said OpenAI repackaged capabilities from the Codex app, released in early 2026, extending its shift from agentic coding tools to ChatGPT Work and now a single, continuous assistant. The name references negative space in OpenAI’s logo, while Elon Musk owns dot.com.
Persistent operation increases OpenAI’s token usage and can surface calendar conflicts, deadlines and technical failures. Timing remains unreliable: an early dot warned Embiricos of a conflict only three minutes beforehand. At a drone delivery startup, however, a dot reportedly detected and analyzed an outage, then proposed fixes before an engineer noticed it. Dots’ proactive research periodically reviews conversations, personal context and authorized resources to suggest help.
Users can disable conversation-based model training and choose accessible resources, but controls lack granularity and dots rejects read-only cloud-drive access. OpenAI launched with conservative defaults and frequent permission requests; Embiricos expects finer controls and better timing. Hark differentiates its assistant by saying it does not collect advertising data as OpenAI and Meta do, while Muse has faced criticism for profiling users’ friends and family. Scaling also requires outside acceptance: Embiricos keeps a human assistant for corporate partners and refuses meetings with other people’s agents. OpenAI is discussing payment access with processors, potentially allowing dots to spend money and raising the trust required from inboxes to bank accounts.
Positives
- $100 monthly gives Pro subscribers or higher access to dots across ChatGPT’s desktop and mobile apps.
- Three-minute conflict warnings are expected to improve as OpenAI refines dots’ proactive timing.
- A drone delivery startup’s dot reportedly analyzed an outage and proposed fixes before an engineer noticed it.
- Proactive research uses authorized resources and conversational context to identify potentially useful actions without prompting.
- Model-training opt-outs and resource-access settings provide users with baseline privacy controls.
Risks & concerns
- Three minutes of notice made one early calendar-conflict warning effectively too late for Embiricos to respond.
- Read-only cloud-drive access is unavailable, forcing users to grant broader permissions or withhold access entirely.
- Current privacy controls lack the granular permissions Embiricos says OpenAI still needs to build.
- Muse’s profiling controversy and Hark’s advertising-data criticism highlight the privacy risks of persistent assistants.
- Payment-processor discussions could extend dots’ reach from personal communications to users’ money.
- Embiricos still relies on a human assistant because corporate partners may resist interacting with AI agents.