Apple's New Siri AI Arrives, and the Early Reviews Are Mixed

Today’s Focus

Apple has begun rolling out the long-promised AI overhaul of Siri, more than a year after the company first previewed the assistant at its 2024 developer conference and then quietly pushed the launch back.

The Verge’s Allison Johnson, who tested the new assistant on a developer build, wrote that the revamped Siri can now pull context from on-screen content, act across apps, and answer multi-step requests that tripped up the old version. Her verdict: “it actually works,” though it is slower and narrower than competitors like ChatGPT or Google Gemini.

The technical foundation was detailed in a paper published this week by Apple Machine Learning Research, which described a new family of “Apple Foundation Models” running both on-device and on Apple’s Private Cloud Compute servers.

Nvidia confirmed in a corporate blog post that its confidential-computing GPUs will help power that cloud tier, a notable concession for Apple, which has historically avoided relying on Nvidia silicon.

The Economist framed the launch as Apple positioning itself as a “dark horse” in generative AI, betting that privacy guarantees and tight operating-system integration matter more to consumers than raw model performance.

Investors disagreed. Yahoo Finance reported that Apple shares fell after the Worldwide Developers Conference (WWDC) keynote, with analysts citing a lack of a clear answer to OpenAI and Google and questions about how Apple will monetize the features. The new Siri is shipping in beta to developers now and is scheduled for a public release with iOS 27 in the fall.

The Debate

Supporters argue

Backers of Apple’s approach say the company is doing AI the right way, even if it is doing it slowly. The Economist argued that Apple’s combination of on-device processing and its Private Cloud Compute architecture offers privacy protections that rivals “cannot easily match,” giving Apple a structural advantage with regulators in the EU and with privacy-conscious users.

The Verge’s review emphasized utility over novelty. Johnson wrote that the new Siri’s ability to take action inside apps, summarize messages, and chain commands together is “the thing I actually want from an assistant,” in contrast to chatbots that produce impressive essays but cannot send a calendar invite.

Apple’s own research team contends that smaller, specialized foundation models tuned for specific tasks can match much larger general-purpose models on the workloads most users actually run. Nvidia, in its blog post announcing the Private Cloud Compute partnership, called Apple’s confidential-computing design “a new bar” for verifiable user-data protection in commercial AI systems.

Critics argue

Skeptics say Apple is still well behind and that this week’s rollout did not change that. Yahoo Finance reported that several Wall Street analysts cut price targets or flagged disappointment after WWDC, with one note saying Apple “showed up to the AI race a year late and still looks winded.”

Critics on the product side argue the new Siri is missing features Apple itself promised in 2024. Bloomberg reporter Mark Gurman, cited across coverage of the launch, has noted that the most ambitious capabilities, including deep personal-context awareness, were pushed into a later release.

Some developers and commentators contend Apple’s reliance on partner models, including a ChatGPT integration disclosed last year and reportedly expanded talks with Google, undermines the company’s “we built it ourselves” framing. The Verge noted that for many complex queries, Siri still hands the request off to OpenAI, raising the question of what users are getting from Apple’s own models that they could not get from a free ChatGPT app.

What the experts say

Independent researchers say the gap between Apple and the leading AI labs is real but narrower than the stock-market reaction suggests. Ethan Mollick, a Wharton professor who studies generative AI adoption, has written that on-device assistants with tool-use capabilities, the category Apple is targeting, remain largely unsolved across the industry.

Benchmark data published by Stanford’s Institute for Human-Centered AI in its 2025 AI Index found that the performance gap between the top closed model and the top open or smaller model shrank from roughly 12 percentage points in early 2024 to under 2 points by late 2025, suggesting Apple’s smaller-model strategy is defensible on technical grounds.

Pew Research Center surveys from 2024 found that 52% of US adults are more concerned than excited about AI in daily life, with privacy ranking as the top concern. That data point cuts in Apple’s favor, but Pew also found that only 33% of respondents trusted any major tech company to handle AI responsibly, including Apple.

By the Numbers

$3.0 trillion: approximate market capitalization of Apple as of the WWDC 2026 keynote, per Yahoo Finance.
1 year: length of the delay between Apple’s original 2024 Siri announcement and this week’s developer rollout, according to The Verge.
2 points: narrowing of the benchmark gap between leading closed AI models and smaller or open models from 2024 to 2025, per Stanford’s 2025 AI Index.
52%: share of US adults more concerned than excited about AI in daily life, according to Pew Research Center.
33%: share of US adults who say they trust major tech companies, including Apple, to develop AI responsibly, per Pew.
iOS 27: the operating system release scheduled to carry the new Siri to the public in fall 2026, per Apple’s developer documentation cited by The Verge.
2: number of model tiers in Apple’s new foundation-model family, one running on-device and one in Private Cloud Compute, per Apple Machine Learning Research.