Apple is taking an unusual swing at one of its browser's most persistent vulnerabilities. The Verge reports that the company is deploying AI tools to let users build their own Safari extensions, essentially lowering the floor on what it takes to create browser add-ons by allowing people to describe what they want in plain language rather than writing code by hand.
To understand why this matters, it helps to understand how Safari arrived at this position. The browser has always occupied a peculiar place in the ecosystem. On iPhone, it is effectively the only real option — Apple's policies have historically required competing browsers to use WebKit, the same rendering engine Safari runs on, stripping rivals of meaningful differentiation. On the Mac, however, Safari competes more honestly against Chrome and Firefox, and that competition has exposed a structural weakness that Apple has never quite solved: the extension library.
Extensions are, for a significant portion of power users, the entire point of a browser. The ability to install an ad blocker, a password manager with deep integration, a read-later tool, or any number of productivity utilities is what separates a browser someone tolerates from one someone prefers. Chrome built its dominance in part on the Google Web Store, which over years accumulated hundreds of thousands of extensions spanning nearly every conceivable use case. Firefox maintained its own loyal constituency by supporting a similarly broad and well-maintained library. Safari's equivalent catalog has always felt sparse by comparison.
The reason for that sparsity is not accidental. Apple imposed notably stricter requirements on Safari extension developers — requirements that made sense from a security and privacy standpoint but created real friction for developers weighing where to spend their time. Porting a Chrome extension to Safari was rarely trivial, and the comparatively smaller Safari user base on the desktop made the economics unattractive. Developers naturally gravitated toward Chrome first, Safari as an afterthought or never at all. Apple made efforts over the years to court developers and streamline the porting process, with mixed results. The gap never fully closed.
The AI-assisted extension builder The Verge describes is a different kind of answer to that problem. Rather than continuing to persuade professional developers to support Safari, Apple appears to be betting that it can expand the supply of extensions by enabling the people who want them to create them. This is the logic sometimes called vibe coding — a term that has gained currency recently to describe the practice of using large language models to generate functional code from natural-language descriptions, without the user needing to understand what the code actually does. Someone who wants a button that strips tracking parameters from URLs, or a floating word-count display for any text field, could theoretically describe that behavior and receive a working extension in return.
The strategic appeal is real. If the bottleneck has always been developer willingness to build for Safari, shifting extension creation to end users sidesteps the bottleneck entirely. The approach also fits Apple's broader posture on AI, which has emphasized on-device, privacy-preserving capabilities and practical utility over more speculative applications. Building a small, sandboxed browser extension is exactly the kind of bounded, concrete task where current AI code generation tends to perform well.
There are limits worth acknowledging. Extensions built this way will almost certainly vary wildly in quality and reliability. A user who cannot evaluate code will not know whether their AI-generated extension is doing something unintended, and the history of browser extensions includes plenty of cautionary examples — malicious or poorly written add-ons that exfiltrated data, degraded performance, or broke without warning. Apple's sandboxing and review architecture may catch some failure modes, but the degree to which these user-generated extensions enter any kind of formal review pipeline is an important detail that will shape how safely this scales.
The likely reading is that this addresses the perception problem as much as the practical one. A user who builds their own workaround, however simple, is a user who stays in Safari rather than switching to Chrome for a particular workflow. Even if the extensions produced are modest in scope, the act of solving a problem inside the browser rather than abandoning it has retention value.
For the broader browser market, the development is worth watching as a potential template. Chrome and Firefox both benefit from their existing libraries, but neither has moved this aggressively to democratize extension creation. If Apple demonstrates that AI-assisted tooling meaningfully expands what users can do without developer mediation, the pressure on competitors to respond will grow.
The immediate question is how widely Apple plans to roll this capability out, under what conditions user-generated extensions can be shared, and whether any quality or safety review applies before they run. Those details will determine whether this is a genuine expansion of Safari's utility or a clever demo that dissolves on contact with everyday use.
