Mark Gurman, Bloomberg’s lead Apple reporter, said on Sunday that he expects Apple to “try to create its own OpenClaw competitor, delivering a system that could fully operate its software across iPhones, iPads and Macs on behalf of the user.” The comment came in a thread following his Bloomberg Power On newsletter reviewing Apple’s overhauled Siri AI features in iOS 27 and macOS 27.
The prediction is not a product announcement. Apple has not confirmed any OpenClaw-style agent project. But Gurman’s track record on Apple product direction makes this more than idle speculation.
The Groundwork Already Exists
Apple has been quietly laying infrastructure. In May, The Information reported (via 9to5Mac) that Apple is “designing a system” for AI agents to operate within App Store privacy and security standards. That effort focused on letting third-party agents interact with apps while maintaining Apple’s control over permissions and data access.
Building its own agent is the logical next step. If Apple is already creating the rules for how agents behave on its platforms, it has every incentive to build the reference implementation.
The Hardware Advantage and the Privacy Problem
Wccftech’s analysis points to Apple’s unified memory architecture as a differentiator. An Apple-built agent running on-device across M-series chips could process tasks without sending data to external servers, a privacy advantage that neither OpenClaw nor Codex can match in their current cloud-dependent architectures.
The tradeoff is capability versus control. As Wccftech notes, if Apple locks down agent permissions too tightly to protect user privacy, the agent becomes less useful than competitors that operate with broader system access. If Apple loosens control and an agent mishandles credentials or deletes files, the PR damage would be severe for a company that has made privacy its core brand promise.
OpenClaw’s model, which requires explicit user permission before file system changes and defaults to locked access, offers one template. But OpenClaw users can also grant unrestricted access, a choice Apple may not be willing to offer.
The Bundling Question
Gurman’s framing implies Apple would integrate its agent into existing subscription tiers rather than selling it standalone. Apple One already bundles iCloud, Music, TV+, Arcade, Fitness+, and News+. Adding an AI agent to that bundle would undercut competitors on price while locking users deeper into the Apple ecosystem.
This is a familiar playbook. Apple Music did not need to be the best music service. It needed to be good enough and bundled. The same logic could apply to an Apple agent: it does not need to match OpenClaw’s flexibility if it runs natively across every Apple device a user owns.
What the Agent Ecosystem Should Watch
Gurman used the word “longer term.” This is not an iOS 27 feature. But the direction matters now for three reasons.
First, Apple controls the operating system. An Apple-built agent would have API access that third-party agents cannot get, from Shortcuts integration to system-level process management. OpenClaw on macOS runs as a user-space application. An Apple agent could run as a first-party service.
Second, the App Store agent framework Apple is building will define what third-party agents can and cannot do on iOS. Those rules will inevitably favor Apple’s own agent, just as App Store rules have historically favored Apple’s own services.
Third, the 2+ billion active Apple devices represent a distribution channel that no agent startup can replicate. If Apple ships a “good enough” agent pre-installed on every iPhone, the addressable market for standalone agent products on Apple platforms shrinks overnight.
None of this is confirmed product. But when Bloomberg’s most reliable Apple source says he expects it, the smart bet is to plan for it.