Apple has released a rebuilt Siri powered by Google's Gemini models after years of development delays. The revamped assistant combines on-device processing with cloud computation but will not launch in the European Union initially.
Apple's new Siri represents a significant overhaul of its virtual assistant, marking the company's shift toward AI-driven capabilities. The system leverages Google's Gemini models and operates through a hybrid architecture: some processing occurs directly on the device, while more complex tasks route through Apple's Private Cloud Compute infrastructure.
Early testers highlight improvements in handling multi-step requests and leveraging on-screen context to deliver more relevant responses. The assistant can now process complex sequences of commands and reference visual information from the user's display, addressing longstanding limitations of the previous version.
However, the rollout reveals current constraints. Users report instances of hallucinations—where Siri generates plausible-sounding but inaccurate information—and gaps in accessing personal context data. These limitations suggest the system requires further refinement before achieving full maturity.
Notably, Siri remains unavailable in the European Union despite the global launch. Apple has not publicly specified a timeline for EU availability, likely due to regulatory compliance considerations under the Digital Markets Act and AI Act, which impose stricter requirements on large technology companies operating in the region.
The partnership with Google marks a notable strategic decision for Apple, which has historically developed AI systems in-house. By building Siri on Gemini, Apple gains access to Google's advanced language models while maintaining control over how data flows through its Private Cloud Compute system.
The launch follows years of delays and signals Apple's determination to modernize its assistant as competition intensifies from ChatGPT, Claude, and other AI-powered tools. Whether the improvements justify the extended development timeline will depend on how effectively Apple addresses the reported hallucinations and context limitations in subsequent updates.
A simulated AI-generated WeChat bug demonstrates why slowing the global AI race requires coordination between the US and China. Bloomberg Opinion columnist Catherine Thorbecke argues that transparency on AI risks must become a shared priority.
Paying for frontier AI models provides only a four-month capability advantage over free open alternatives, according to a Mozilla report. The premium pricing doesn't justify the minimal head start for most users.
AI-generated tributes and conversational bots designed to simulate deceased people are emerging as new mourning tools. Experts warn the trend could increase isolation during grief.
New York State Assemblyman Alex Bores has launched Who Decides, a nonprofit organization that has already raised $10 million to influence Democratic artificial intelligence policy, with plans to secure an additional $20 million by 2027.