The Siri Admission: When Apple Acknowledged Its AI Models Couldn’t Compete

The most damning evidence of Apple’s AI struggles came from Bloomberg’s Mark Gurman: Apple is considering using AI technology from Anthropic or OpenAI to power a new version of Siri, sidelining its own in-house models.

The Internal Testing Reality

Apple held a head-to-head “bake-off” competition:

• Tested models on complex user queries
• Compared performance head-to-head
• Google started training custom models for Apple’s servers

The Results

Apple’s models lost consistently across all tests

The Painful Finding

“Third-party models — particularly Claude — outperformed Apple’s tools in complex queries”

• $34.5B Annual R&D
• 2+ years behind competitors
• FAIL on internal models

What This Means

1. R&D Didn’t Deliver — $34.5B annual spend, Apple couldn’t build competitive AI models
2. Vertical Integration Broken — The control-everything approach that built iPhone doesn’t work for AI
3. Competitor Dependency — Now needs Google/OpenAI/Anthropic to power core product intelligence
4. Brand at Risk — Siri updates promised in 2024, still undelivered in 2026

The Key Quote

“sidelining its own in-house models”

Apple publicly admitting defeat on AI development — unprecedented for a company built on control.

This is part of a comprehensive analysis. Read the full analysis on The Business Engineer.