TLDR
Apple is hiring a senior engineering manager to lead an AI agent that can operate a Mac.
The project is in early stages, with a possible launch as soon as late 2027.
The agent would browse the web, use apps, run commands and write code.
macOS 27.3 will warn users against giving full access to third-party agents.
Apple says broad Full Disk Access poses growing risks as AI agents become more capable.
Apple is preparing new warnings for Mac users who give outside AI agents full access to their computers. At the same time, the company is hiring to build an AI agent of its own that can operate a Mac.
Bloomberg reported on November 19 that the upcoming macOS 27.3 update will show warnings that discourage users from granting full access to third-party agents. It also reported that a recent Apple job listing describes an AI agent being built inside the company.
The listing seeks a senior engineering manager to oversee the effort. The project is still in its early stages.
The listing points to an agent that could handle many kinds of computer work on a Mac. Apple has not shown any version of the tool in public.
New Limits on Outside Agents
Apple first raised its concerns in an October 2 notice to developers. It said some developers use Full Disk Access in ways that could expose files, mail, messages and browsing history without users fully knowing or understanding.
Full Disk Access largely bypasses the normal privacy controls on a Mac. The permission exists so that tools such as backup apps can work properly.
For communication apps, this level of access can also affect people who exchange messages with the user. A single permission can reach beyond the person who granted it.
View post by @PatentlyApple on X
Apple plans to add controls so that granting this access requires very explicit action from the user. The company tied the change to risks that are growing quickly as AI agents become more capable and act more on their own.
The goal is to make sure users understand those risks before they make choices about their data and privacy. The macOS 27.3 warnings follow that same direction.
What Apple’s Own Agent Would Do
The job listing lays out a wide set of tasks for Apple’s agent. It would browse the web, work across apps, run commands, write and debug code, and call outside services.
The agent would also pick the right tool for each step by itself. The listing describes agents moving between apps, browsers, command lines, code and APIs to finish tasks with several steps.
Private operation and the polish found in other Apple products are part of the stated goals. These are aims written in a hiring post, not features shown in a working product.
Anthropic and OpenAI have already launched services that can do work on a computer. Meta has also recently introduced its Muse line of agents, which includes a Mac app with access to files, messages and calendars.
An Apple agent would bring that competition directly onto the Mac. There, access to local apps, files and system permissions shapes what any agent can do.
Those are the same areas where Apple is now adding limits for outside software. Under the coming changes, outside agents seeking broad access will face extra steps and warnings.
Apple has looked at launching its agent service as soon as the end of 2027. That date is a reported possibility, and the company has not announced a release.

