OpenAI Codex Locked Use
OpenAI Codex Can Now Control Your Mac Even When Locked
OpenAI's Codex desktop agent for Mac can now operate applications and complete tasks even after the screen is locked — a capability the company calls "Locked Use." The feature, announced May 21, 2026, uses an Apple authorization plug‑in that temporarily unlocks the Mac with strict temporal and behavioral safeguards, letting developers trigger and monitor long‑running agent tasks remotely from their phone. The update also shipped Appshots for instant window context, graduated Goal Mode to general availability, and improved the in‑app browser.
Codex Breaks Through the Lock Screen
One of the longest‑standing frustrations for AI agent users on macOS has been the lock screen. Leave your Mac unattended during a multi‑hour coding task, and the screen locks — abruptly halting whatever the agent was doing. On May 21, 2026, OpenAI shipped a solution: Codex can now continue operating your Mac even after the screen is locked and the display goes dark.
The feature, called "Locked Use," was announced by the @OpenAIDevs account on X: "Codex anywhere and everywhere, all the time. Now your Mac doesn't have to be unlocked for Codex to use your computer. From your phone, Codex can securely use apps on your Mac, even when the screen is off and locked." The post quickly racked up 1.8 million views, with prominent developer voices like Theo Browne (@theo) remarking: "At this point I think OpenAI has more MacOS engineers than Apple does."
How Locked Use Works Under the Hood
According to OpenAI's developer documentation, Locked Use is not a hack or a bypass — it integrates directly with macOS's trusted security architecture. When you enable the toggle in Codex settings, the app installs an Apple authorization plug‑in that participates in the standard macOS unlock flow. When Codex needs to access an application via its Computer Use feature after the Mac has locked, the plug‑in temporarily unlocks the machine while blocking local use and preserving the locked‑screen protections.
The implementation runs through an Apple authorization plug‑in that participates in the macOS unlock flow. As byteiota notes, this is not a bypass — it works within Apple's existing security framework rather than circumventing it.
Security by Design — The Safeguard Layer
OpenAI has baked multiple overlapping safeguards into Locked Use. As Macworld's Roman Loyola reports, these include: short‑lived authorization — the unlock window is scoped to the current Codex task and expires quickly; exclusive access — only Codex can trigger the unlock, no other apps or local processes can use it; full‑display coverage — Codex covers every connected display while the desktop is temporarily unlocked, preventing anyone nearby from seeing screen content; and auto‑relock on local input — if the system detects any physical keyboard or pointer activity, the Mac immediately re‑locks and pauses automatic unlock until a manual unlock occurs.
OpenAI's documentation states bluntly: "Locked use is intentionally narrow. It's not a general‑purpose remote‑unlock path for your Mac, and it doesn't let other apps or local processes unlock the computer."
What Locked Use Can't Do
The feature has a few important constraints. Most notably, Locked Use does not work when you close your MacBook's lid — clamshell mode triggers a fundamentally different sleep path that the authorization plug‑in cannot intercept. The feature is also macOS‑only (no Windows or Linux support) and is unavailable in the European Economic Area, the United Kingdom, and Switzerland at launch, reflecting regional regulatory constraints on automated desktop access.
Additionally, Locked Use only works with Codex's Computer Use feature — it does not grant remote desktop access, and the remote control pathway flows through Codex Mobile on iOS or Android via the ChatGPT mobile app preview launched May 14, which syncs live state across devices through a secure relay layer.
A Broader Update: Appshots, Goal Mode, and More
Locked Use shipped as part of a larger Codex update on May 21 — internally dubbed "Codex Thursday" at OpenAI,.1 The release also introduced Appshots, a macOS feature that lets users press Command‑Command to instantly inject the frontmost window — including text beyond the visible scroll area — into any Codex thread. 9to5Mac's Zac Hall notes that Appshots captures "both a screenshot and text from the window, including content beyond what's visible onscreen."
Goal Mode graduated from experimental to general availability across the Codex app, VS Code and JetBrains extensions, and CLI. It lets developers assign a broad objective — such as increasing test coverage to 90% in a specific module — and Codex will plan, execute, check its output, and course‑correct across session breaks until the goal is met. The in‑app browser received a precision upgrade with an annotation mode, and Business and Enterprise users gained shared plugin capabilities.
Why Locked Use Matters for Agentic AI
The Locked Use feature represents a meaningful step in the evolution from AI assistants that require constant human presence to agents that operate autonomously in the background. Previously, developers running long Codex tasks resorted to workarounds like caffeinate terminal sessions to prevent sleep, third‑party utilities to disable lock‑screen triggers, or hardware display dongles that trick the Mac into thinking an external monitor is attached. Locked Use eliminates these hacks with a native, Apple‑integrated solution.
Federico Viticci of MacStories called Codex's Computer Use "the best computer use feature I have ever tested in any LLM or desktop agent," praising its use of the macOS accessibility hierarchy (AX Tree) rather than the screenshot‑and‑coordinate‑click approach used by competitors. This lets Codex understand the full structure of UI elements — buttons, text fields, nested menus — giving it more precision than Claude or Perplexity's desktop agents. When paired with Locked Use and Goal Mode, Codex can now run substantial overnight workflows without a human ever touching the keyboard.
The timing aligns with OpenAI's push to make Codex an always‑available collaborator. With over 4 million weekly users, the combination of mobile access, goal‑oriented task execution, and locked‑screen operation turns Codex from a desktop‑bound tool into a persistent agent that works even when you're not there.
How to Enable Locked Use on Your Mac
Enabling Locked Use takes a few steps. First, you need the Codex desktop app for macOS with the Computer Use plugin installed. Then:
- Open Codex Settings → Computer Use.
- Toggle on "Locked computer use" — this installs the Apple authorization plug‑in.
- Verify that Codex has both Screen Recording and Accessibility permissions in System Settings → Privacy & Security.
- Start a Computer Use task from a connected device (your phone via the ChatGPT mobile app) after your Mac's screen has locked.
When you're ready to run a long task, you can lock your Mac normally, and Codex will continue working. The screen will remain covered (showing the lock screen) while Codex operates apps in the background. If you return and touch the keyboard or trackpad, the Mac immediately re‑locks for security.
The feature is available now to all Codex users on macOS, though as noted, it is not yet available in the EEA, UK, or Switzerland. Full documentation is at developers.openai.com/codex/app/computer‑use.
Sources
- 1.byteiota(byteiota.com)
Related News
May 23, 2026
Anthropic and OpenAI Race to Embed AI Agents on Wall Street
Within a 72-hour window in May 2026, Anthropic and OpenAI each launched enterprise deployment arms, announced major financial-services partnerships, and shipped agent tooling targeting Wall Street's most critical workflows. The race to become the operating system for finance is accelerating — and the stakes have never been higher.
May 22, 2026
Pentagon Tests OpenAI and Google AI to Replace Anthropic's Claude
The Pentagon is testing AI models from OpenAI and Google to replace Anthropic's Claude, after Defense Secretary Hegseth designated the company a supply-chain risk. The tests began in early March with 25 'power users' across the department, while Anthropic battles the designation in court arguing it could cost billions.
May 22, 2026
Intuit Lays Off 17% of Workforce as AI Restructuring Wave Spreads
Intuit is cutting about 3,000 jobs — 17% of its workforce — while simultaneously signing multi-year AI deals with Anthropic and OpenAI. The maker of TurboTax, QuickBooks, and Mailchimp joins Meta, Amazon, and Block in a wave of 2026 layoffs where AI investment and headcount reduction go hand in hand.