New report reveals chaos behind Apple's AI push as Siri overhaul falters

A revealing investigation from The Information has pulled back the curtain on the chaos surrounding Apple’s efforts to transform Siri into a true AI-powered assistant. What was announced with confidence at WWDC 2024 as “Apple Intelligence” is, behind the scenes, the product of years of internal dysfunction, shifting strategies, and growing discontent among Apple’s own engineers.

Rather than a clear path forward, Apple’s attempt to reimagine Siri has been marked by indecision and frequent changes in technical direction. According to the report, teams were initially instructed to pursue a hybrid approach with separate small and large language models—internally dubbed “Mini Mouse” and “Mighty Mouse.” But the plan was scrapped in favor of a single, cloud-based model. This sudden pivot was one of many, leading to delays, burnout, and the departure of key talent.

The turmoil wasn’t just technical—it was cultural. More than half a dozen former employees told The Information that Apple’s AI and machine learning division was riddled with poor leadership and lacked urgency. Nicknamed “AIMLess” internally, the division became known for passing Siri around between teams without any long-term ownership or progress.

Engineers described an environment where ambition was stifled, risk-taking was discouraged, and focus was often placed on small, low-impact wins like removing the “Hey” from “Hey Siri”—a change that reportedly took over two years.

Siri's troubled development also created tension across teams, especially as competitors like OpenAI rapidly redefined expectations for conversational AI. Apple executives were initially slow to respond to ChatGPT’s rise, with AI chief John Giannandrea allegedly downplaying its significance. Meanwhile, Apple's own models failed to keep pace in internal benchmarks, and efforts to use more advanced emotional intelligence or general knowledge scraping were either blocked or ignored.

One of the most damning details in the report is that the impressive Siri features shown at WWDC—like contextual awareness across apps and dynamic real-time suggestions—were not functional on actual test devices. Engineers were surprised to see these capabilities demonstrated at all, with the only working feature on internal builds being the animated AI ribbon around the screen.

Despite this, there may be a shift underway. Sources say Craig Federighi has begun pushing for a more aggressive strategy, including potential use of third-party or open-source models if needed. The directive now is to deliver real AI value—regardless of whether it comes from within Apple or outside it.

The Siri project, once symbolic of Apple’s lead in voice technology, is now a reminder of how quickly the company fell behind in AI. As pressure mounts from competitors and users alike, Apple’s next steps will need to prove it can move past the internal dysfunction and finally bring its assistant into the modern AI era.