Apple greenlights third-party AI models to save Siri from falling behind

After years of strict internal limitations, Apple is finally allowing its engineers to integrate third-party AI models into Siri—marking a dramatic change in strategy as the company fights to keep pace in the rapidly evolving AI race. In a new report from The Information, the decision, driven by Software Chief Craig Federighi, breaks from Apple’s long-standing policy of relying solely on homegrown technology, even when it fell short.

Inside Apple, this move represents a turning point. Engineers who had previously tested powerful language models from companies like OpenAI were told they could only use them for benchmarking, not real-world features. That restriction caused deep frustration among teams tasked with building new Siri capabilities, especially as Apple’s internal models repeatedly failed to match the sophistication of their external counterparts.

The shift in policy empowers teams to use whatever tools necessary—open-source or commercial—to deliver meaningful, next-gen Siri experiences. Federighi’s directive contrasts sharply with the approach of John Giannandrea, Apple’s AI lead, who kept the focus on developing in-house models despite performance limitations that slowed progress across Siri’s roadmap.

For Apple, which has historically avoided external dependencies in favor of control and privacy, this is a significant change in philosophy. But with competitors like Google, Microsoft, and OpenAI accelerating AI innovation, the pressure to modernize Siri has become too great to ignore.

Rather than waiting for internal tech to catch up, Apple is choosing pragmatism. The focus is now on delivering the best AI-powered user experience possible—even if it means breaking tradition and borrowing from the outside. For users, it could finally bring the Siri overhaul Apple has long promised but never fully delivered.