Apple is taking major steps to rebuild Siri from the ground up after a rocky rollout of its AI-powered features under “Apple Intelligence.” According to a detailed report by Bloomberg’s Mark Gurman, the company is now focusing on creating a new version of Siri based entirely on large language models (LLMs), internally called “LLM Siri,” in an effort to catch up in the AI race.
The original integration of generative AI features into the legacy Siri system proved ineffective. Apple employees likened the process to a game of “whack-a-mole,” where fixing one problem led to several new ones. Several internal challenges reportedly contributed to the setbacks. Software head Craig Federighi was hesitant to invest heavily in AI without a clear direction, delaying the purchase of powerful GPUs necessary for development. Additionally, Apple began exploring AI seriously only after ChatGPT launched in late 2022, putting it behind competitors like OpenAI and Google.
AI chief John Giannandrea, who joined Apple in 2018 from Google, believed users preferred to disable AI chat tools. He reportedly lacked influence within Apple’s leadership circle and didn’t push aggressively for resources. As a result, Apple’s AI capabilities remained limited, while marketing overpromised features like a smarter Siri that could understand context across apps — features that were later postponed.
To fix this, Apple is now developing a fully LLM-based Siri through its AI team in Zurich. The goal is to create an assistant that can engage in natural conversations and better understand user requests. The company is also working on privacy-centric ways to improve its AI models using on-device data from iPhones. Part of the future strategy may involve turning Siri into an AI-powered web search assistant, potentially in partnership with companies like Perplexity.
