A look at Craig Federighi's approach to AI at Apple; sources say he led a push to use other companies' models to help launch a revamped Siri later in 2026
The Information:
Context & Ripple Effects
Apple's reported external-model push follows an earlier evaluation of outside models for Siri, in which Anthropic's technology was said to be the most promising. It marks a sharper operational response to the gap between Siri and rival assistants described across the coverage.
The reporting also follows concerns that personalized Siri might need to be rebuilt from scratch. Putting Mike Rockwell and Siri under Federighi, as described in the corpus, concentrates both product accountability and the model-selection decision in one organization.
First-order effects
- Apple's Siri team would shift from relying solely on internally developed AI toward integrating third-party models for the planned 2026 revamp, if the reported approach is adopted.
- Federighi's organization becomes the immediate decision-maker for model evaluation, product integration, and the trade-offs between outside capabilities and Apple's assistant experience.
Second-order effects
- External model providers gain a route to compete for distribution inside Siri, while Apple gains leverage to compare suppliers rather than tying the assistant's progress to a single in-house model path.
- A more capable Siri would raise the bar for competing assistants on the iPhone, especially where the existing product is described as needing to catch up; the earlier external-model testing effort suggests that comparison process is already underway.
Third-order effects
- If this pattern persists, the assistant market may increasingly separate model development from device-level distribution: platform owners can assemble assistants from multiple model suppliers while retaining the customer interface.
- That shift would make integration quality, privacy controls, and the ability to switch or combine models more important sources of platform advantage than exclusive ownership of every underlying model.
The trend: This is one data point in the shift toward platform companies treating frontier AI models as interchangeable inputs while competing to own the assistant distribution layer.