OpenAI Develops Codex Control for Locked Macs as May 14 Remote Feature Expands

4 articles · Updated · TestingCatalog · May 16

OpenAI is developing a Codex capability that would let its agent operate macOS apps even when a Mac is locked or asleep, extending remote control beyond the current unlocked-session limit.
Computer Use has been the main blocker because it needs a visible, awake desktop to move the cursor, type in apps and read the screen, forcing users back to the machine to log in.
The feature would let a phone trigger tasks such as opening desktop apps, testing GUI builds, running simulators or pulling from data sources without physically waking the Mac.
May 14 marked the first step, when ChatGPT's mobile app added Codex remote controls for iPhone and Android users to review outputs, approve commands, switch models and send new tasks to a Mac.
OpenAI is also exploring control of multiple Codex devices, including machines like a Mac Mini, though any locked-screen access could face scrutiny from Apple's macOS security model.

Can OpenAI's agent truly control a locked Mac without creating an unavoidable backdoor for hackers?

With AI discovering kernel exploits, will Apple permit agents to operate on locked devices?

Is letting AI work on our locked computers the next leap in productivity or a catastrophic security failure?