OpenAI has begun rolling out native messaging capabilities for its autonomous Dot personal agents, enabling select U.S. Pro subscribers to assign tasks, send photos, and exchange voice memos directly through iMessage and Rich Communication Services (RCS). But it may come at a personal cost.
The rollout fulfills a pledge OpenAI made during the artificial intelligence (AI) agents’ debut last month to detach them from the standalone ChatGPT interface.
Previously confined to mobile, desktop, and web applications, Dots can now initiate proactive outbound contact to deliver project updates directly to a user’s native text feed. The integration effectively extends the agent’s reach to smartwatches and system-level notifications without requiring dedicated client software.
However, the convenience introduces a notable trade-off in data privacy.
To bridge ChatGPT with native telecommunications networks, OpenAI partnered with third-party messaging infrastructure provider Linq. Because Linq acts as an intermediary relaying and processing transmissions before they reach OpenAI’s servers, interactions are not end-to-end encrypted — even when channeled through Apple Inc.’s iMessage or Google’s encrypted RCS protocols.
OpenAI acknowledged the architectural limitation in technical documentation, stating that while it conducted “extensive penetration testing” with internal frontier models and independent cybersecurity firms, the routing design precludes true end-to-end encryption.
The company stressed that security controls restrict communications exclusively to verified one-to-one channels: outside parties cannot text a user’s Dot, nor can the agent message anyone else or enter group threads.
Onboarding requires users to navigate to their Dot profile across the desktop or mobile app, toggle the “Texting” option, and authenticate a domestic phone number. Legacy SMS is explicitly unsupported; messages sent via standard cellular protocol trigger an automated redirect. Subscribers can sever the connection at any time by texting “STOP” or disabling the feature via settings, though disconnecting does not purge conversation histories or memory caches stored within ChatGPT.
The release comes amid intensifying competition across the autonomous assistant sector.
Last week, rival Anthropic unveiled Claude Duet, an agent slated for inclusion in its $20-per-month Pro tier currently restricted behind a domestic waitlist. OpenAI, by contrast, has reserved Dot texting exclusively for higher-paying Pro tier accounts, bypassing standard $20-per-month Plus subscribers, as well as Enterprise and Business workspaces.
By embedding contextual memory across both conversational texting and the core web app, OpenAI is betting that frictionless cellular access will cement user retention, provided enterprise and privacy-conscious consumers are willing to overlook the absence of end-to-end encryption.
