Tim Cook is now Apple’s executive chair, not its CEO. John Ternus succeeded him on September 1, 2026, after Apple announced the transition in April. The AI record examined here largely belongs to Cook’s CEO tenure: Apple presented a more personal, cross-app Siri in 2024 and did not ship the central features on the expected schedule. It then changed AI leadership, partnered with Google for Gemini technology, and introduced a redesigned Siri AI in June 2026. As of September 13, that new system remained a beta rather than a fully proven global rollout.

This sequence is more accurate than calling Apple absent from AI or declaring the crisis solved. Apple shipped other Apple Intelligence features and built an on-device and private-cloud architecture. Its hardest assistant promise arrived late, and availability remains restricted by hardware, language, region, and regulation.

This article was checked on September 13, 2026. Apple announcements are primary sources for product design, availability, and leadership changes. Associated Press reporting provides independent chronology for the delay and the Gemini agreement. Apple’s Form 10-K supplies the regulated business and risk context.

The 2024 promise exceeded the delivered product

At WWDC 2024, Apple introduced Apple Intelligence and demonstrated a Siri that could use personal context, understand what was on screen, and take actions across apps. The launch announcement also described on-device processing, Private Cloud Compute, and optional access to ChatGPT for some requests.

Several Apple Intelligence components shipped in stages, including writing, image, notification, and interface features. The more ambitious Siri capabilities did not arrive on the original timetable. The distinction matters because Apple had used those functions to market supported devices before customers could use them.

Associated Press reporting on WWDC 2025 described Apple pulling back promotion of the delayed features and announcing more incremental updates. The AP account provides an external record of the gap between demonstration and delivery. It does not reveal the complete internal technical cause.

Leadership changed after the miss

Apple announced in December 2025 that John Giannandrea would step down as senior vice president for Machine Learning and AI Strategy, act as an adviser, and retire in spring 2026. The official succession release named Amar Subramanya as vice president of AI reporting to software chief Craig Federighi and distributed other parts of the organization across Apple leaders.

The announcement establishes the new structure and timing. It does not publish a postmortem or say one executive alone caused the Siri delay. Cook is accountable for the company’s product claims and executive design, while model, software, services, operations, and marketing teams made the underlying decisions.

Reorganizing can clarify ownership, but it is not evidence of recovery by itself. The new team still had to produce a system that met quality, latency, privacy, and regional requirements.

Apple added external model technology

In January 2026, the Associated Press reported that Apple and Google had agreed to use Gemini technology as part of the foundation for a more personalized Siri. The AP report describes the arrangement as a multi-year collaboration and places it in the context of Apple’s delayed overhaul.

The agreement weakens a simple vertical-integration story. Apple continues to build its own models and silicon, but it can use a partner when external capability helps close a gap. That is a supply decision, not evidence that all Apple research failed or that Google operates the entire assistant.

Public sources reviewed here do not disclose the full commercial terms, exact routing rules, model versions used for every task, or a complete allocation of data between Apple and Google systems. Those details should not be inferred from the existence of the partnership.

Siri AI reappeared in 2026 with a new architecture

Apple introduced Siri AI on June 8, 2026. Its product release describes personal-context search across messages, mail, and photos; on-screen awareness; broader web knowledge; and actions across applications.

The June release said the features were available for developer testing and would reach users as a beta later in 2026. That beta status is evidence of a product milestone rather than proof of mature availability, reliability at scale, or sustained user adoption.

Apple supplies demonstrations and capability descriptions, but no independent failure-rate study in the cited release. The practical test will include correct retrieval, action selection, permissions, latency, error recovery, and behavior across different devices and languages.

Availability remains uneven

Apple’s June release says Siri AI would not be available in China while the company worked through regulatory requirements. Apple separately said iPhone and iPad availability in the European Union would be delayed, while macOS and visionOS access would follow a different path.

Those are Apple statements about current product plans, not neutral legal conclusions about what the Digital Markets Act requires. The European Commission’s interpretation and any final compliance path may differ from Apple’s framing. This article therefore describes the availability result without adopting Apple’s political characterization as fact.

Regional gaps matter commercially. A global enterprise cannot treat a feature as one uniform service when capability, data processing, or release timing varies by country and device.

Privacy is an architecture, not a complete outcome

Apple designed Apple Intelligence to run some requests on device and send harder work to Private Cloud Compute. This can reduce the amount of personal data handled by a general cloud service and gives Apple a differentiated engineering boundary.

It also creates integration work. The system has to route requests, minimize data, verify server software, preserve context, and return consistent behavior across hardware generations. A privacy goal can constrain model size and serving choices without explaining every schedule decision.

Customers should distinguish architecture claims from verified outcomes. They need to know which requests leave the device, what logs or identifiers persist, how partner models are invoked, and what happens when a requested action is unsupported. A high-level privacy promise cannot answer every workflow question.

Apple’s financial scale does not answer product quality

Apple’s fiscal 2025 Form 10-K describes a business built around devices and services and identifies competition, regulation, security, supply, and new-technology execution as risks. The SEC filing does not break out Apple Intelligence revenue or provide a standalone Siri profit measure.

This limits financial inference. Strong iPhone or services revenue cannot prove that users adopted a particular AI feature, just as an AI delay does not by itself establish a decline in the whole business. Product evidence needs more specific measures.

Useful measures would include eligible-device adoption, repeat use, successful task completion, correction and escalation rates, latency, privacy incidents, and regional availability. Apple has not published all of these for Siri AI in a form that permits independent evaluation.

Cook’s legacy turns on claim-to-delivery discipline

Cook’s operating record gives Apple enormous distribution, custom silicon, a large developer base, and control over the hardware-software interface. Those assets make a personal assistant easier to distribute once it works. They do not make model or agent integration automatic.

The 2024 delay exposed a management problem: marketing confidence moved ahead of product readiness. The 2025 leadership change and Google partnership show a willingness to alter organization and sourcing. The 2026 announcement shows a new technical attempt. Cook retains board-level influence as executive chair, but Ternus now owns the CEO-level execution. Full judgment requires evidence from the beta and later production use.

For Apple, success is not winning a chatbot benchmark in isolation. It is delivering useful actions across personal data and applications without compromising user control or making promises before the system is ready. That is the standard by which Cook’s AI strategy should be measured.

Source note

Sources were checked on September 13, 2026. Apple pages document its CEO transition, product announcements, architecture, availability, and AI leadership change. Associated Press provides independent reporting on the delay and Google agreement. The SEC filing provides company-wide financial and risk context. Siri AI remained a beta, not a fully verified global release.