Apple is considering the integration of AI models from Anthropic or OpenAI to improve its virtual assistant, Siri.

Sources indicate that discussions have occurred regarding training these models on Apple’s cloud infrastructure for testing.

If finalized, this shift would represent a departure from Apple’s current reliance on its in-house Apple Foundation Models.

These models have encountered delays and challenges.

Mike Rockwell, head of Siri, and software engineering leader Craig Federighi are leading the evaluation of external AI models.

After testing various technologies, Anthropic’s Claude has emerged as a potential candidate for Siri, though no final decision has been reached.

Despite using ChatGPT for some web-based Siri queries, Apple still relies on its own models for Siri’s core functions.

Negotiations may stall due to Anthropic’s high fee, while internal staff express concern over depending on third-party tech.

🔗 Source: Bloomberg

🧠 Food for thought

1️⃣ Apple’s historical pendulum between control and capability

Apple’s potential partnership with OpenAI or Anthropic represents a rare departure from its long-standing preference for in-house technology development.

This tension dates back to the 1990s when Apple began exploring machine learning with handwriting recognition on the Newton PDA, establishing a pattern of prioritizing vertical integration 1.

When Siri transitioned to a neural network-based system in 2014, Apple made the deliberate choice to keep development internal despite the significant engineering challenges, maintaining control over both hardware and software 2.

Apple’s history shows that when facing significant technological gaps—like when they acquired NeXT in 1996 to modernize their operating system—they’re willing to make pragmatic shifts from their control-centric approach 3.

The current AI inflection point mirrors past moments when Apple had to choose between maintaining complete control or embracing external technology to remain competitive.

Recent Apple developments