For more than a decade, Siri was the assistant Apple seemed almost embarrassed to talk about. At WWDC 2026 that changed. On the keynote stage at Apple Park on June 8, the company retired the cautious, half-finished promises of recent years and introduced a fully rebuilt assistant called Siri AI — and, in a move few would have predicted even two years ago, much of its new intelligence runs on a custom version of Google's Gemini.
What Happened
Apple used the opening of its 2026 Worldwide Developers Conference, held June 8–9, to unveil what it described as the most significant overhaul of Siri since the assistant first shipped in 2011. The redesigned product, now branded simply "Siri AI," is built on a new generation of Apple's on-device foundation models working alongside a large cloud model. It can read on-screen context, understand personal information stored across a user's apps, and chain together actions across multiple applications with far fewer explicit commands than the old Siri required.
The most consequential detail was not the rebrand but the engine underneath it. According to reporting around the event, the cloud layer of Siri AI is powered by a roughly 1.2-trillion-parameter-class Gemini model that Google has customized specifically for Apple, running on infrastructure arranged to keep user data walled off behind Apple's privacy controls. Apple framed the architecture as a hybrid: smaller third-generation Apple Foundation Models handle everyday, latency-sensitive requests on the device, while the heavier reasoning is handed to the customized Gemini system in the cloud. Apple also confirmed that Siri is finally getting its own standalone app rather than living only as a system-level shortcut.
The announcement carried an unusual weight because of who delivered it. WWDC 2026 was widely understood to be Tim Cook's final marquee keynote as chief executive, with the company having signaled a leadership transition later in the year. That framing turned a product launch into something closer to a statement of direction for the next era of the company.
Why It Matters
For most of the modern AI cycle, Apple has been treated as the conspicuous laggard among the largest technology companies. Its rivals shipped capable chatbots and assistants while Apple's own "more personalized Siri," promised at WWDC 2024, slipped repeatedly. Choosing to license a frontier model from Google rather than wait for its in-house systems to catch up is therefore a meaningful admission — and a meaningful strategy. It tells the market that Apple now values getting a competitive assistant into hundreds of millions of hands quickly more than it values owning every layer of the stack itself.
The arrangement is also a quiet reshaping of the competitive map. Google and Apple are direct rivals in phones, browsers, and increasingly in AI, yet the two companies already have one of the most lucrative commercial relationships in the industry through the default-search agreement that sends Google billions of dollars a year. Layering an AI model deal on top of that — reported to be worth on the order of a billion dollars annually — deepens a dependency that regulators on both sides of the Atlantic have been scrutinizing. It also raises the stakes for every other model provider, because the single largest consumer hardware platform in the world has just chosen its partner.
There is a strategic logic for both sides. Apple gets frontier-grade capability without betting its release schedule on internal research timelines. Google gets enormous distribution for Gemini and a validation stamp from the company most associated with consumer trust and privacy. The risk for Apple is the one it has historically refused to accept: handing a core experience to an outside supplier whose incentives are not perfectly aligned with its own.
Reaction
The developer audience that fills WWDC each year reacted with a mix of relief and caution. Relief, because a genuinely capable assistant unlocks the on-screen and cross-app features that third-party apps have wanted to plug into for years. Caution, because a Siri that can take actions across apps raises immediate questions about how much control developers and users will have over what it touches.
Outside the hall, the reaction split along familiar lines. Investors and analysts largely read the move as pragmatic and overdue, noting that shipping something competitive matters more right now than the optics of relying on a rival. Privacy advocates were more measured, pressing Apple to be specific about exactly which requests leave the device, what Google can and cannot see, and how the data boundary is enforced in practice rather than in marketing language. Commentators also noted the symbolism of Apple, long the loudest critic of data-hungry AI, leaning on the company whose business was built on exactly that.
What's Next
The immediate question is execution. Apple has promised an ambitious Siri before and failed to deliver on schedule, so the credibility of this launch will rest on whether Siri AI ships broadly and works as shown when the new operating systems reach the public later in the year. A staged rollout, starting with newer devices capable of running the larger on-device models, is the likely path.
The second question is leadership. With a chief-executive transition expected to follow this keynote, Siri AI becomes an inheritance as much as a launch — the first defining product of whoever steps into the role next. The strategy of pairing Apple's silicon and privacy posture with a best-available external model could become the template for how the company approaches AI broadly, or it could be revisited once Apple's own models mature. Either way, the dependency on Google is now a structural feature of Apple's AI plan, not a temporary patch.
Closing Thoughts
Siri AI is less a single feature than a thesis about how the AI era will actually be distributed. The assistants that reach the most people will not necessarily be the ones with the most impressive demos, but the ones embedded in the devices already in people's pockets. By accepting help from a competitor to get there, Apple has signaled that, at least for now, presence beats purity.
The deeper story is what it reveals about the balance of power in AI. A few model builders are becoming the foundation that even the world's most self-reliant hardware company now stands on. That concentration brings obvious benefits in speed and capability, and equally obvious questions about competition, resilience, and who ultimately controls the intelligence layer of everyday computing. WWDC 2026 will be remembered not just for giving Siri a second life, but for making that dependency impossible to ignore.
한글 요약
애플은 6월 8~9일 열린 WWDC 2026에서 음성 비서를 전면 재설계한 '시리 AI(Siri AI)'를 공개했습니다. 2011년 첫 출시 이후 가장 큰 개편으로, 화면 내용 인식과 앱 간 연속 작업 수행이 핵심입니다. 주목할 점은 클라우드 추론을 구글이 애플 전용으로 맞춤화한 약 1.2조 파라미터급 제미나이(Gemini) 모델이 담당한다는 사실입니다. 가벼운 요청은 기기 내 애플 자체 모델이, 무거운 추론은 맞춤형 제미나이가 처리하는 하이브리드 구조입니다.
이번 발표는 그동안 AI 경쟁에서 뒤처졌다는 평가를 받아 온 애플이, 자체 모델 완성을 기다리기보다 경쟁사의 최신 모델을 도입해 빠르게 시장에 대응하겠다는 전략적 선택으로 읽힙니다. 보도에 따르면 이 협력은 연간 약 10억 달러 규모로 전해지며, 이미 검색 기본값 계약으로 얽힌 두 회사의 의존 관계를 한층 깊게 만듭니다. 개발자들은 앱 연동 기능에 기대를 보이면서도 권한·프라이버시 문제를 지적했고, 투자자들은 대체로 현실적인 결정으로 평가했습니다.
WWDC 2026은 팀 쿡 CEO의 사실상 마지막 주요 기조연설로 여겨져, 시리 AI는 곧 이어질 경영진 교체 이후 차기 리더가 물려받을 첫 핵심 제품이 됐습니다. 관건은 실제 출시 완성도입니다. 애플이 과거에도 시리 개선을 약속하고 일정을 미룬 전례가 있는 만큼, 올해 하반기 새 운영체제와 함께 약속한 기능이 제대로 구현되는지가 신뢰의 시험대가 될 전망입니다. 출처: macobserver, Business Standard, CNBC.