Apple’s artificial intelligence strategy has entered a new phase, and the shift is more fundamental than most users realize. According to Bloomberg reporter Mark Gurman, one of the most connected voices on Apple’s internal operations, the company now “runs on Anthropic” for large portions of its AI stack. The phrase is striking not because it implies Apple abandoned its own work, but because it confirms a deeper transformation: Apple is no longer trying to build everything alone. Instead, it is blending its in-house systems with best-in-class external models to accelerate the next generation of Apple Intelligence and Siri.
This change arrives alongside Apple’s newly confirmed partnership with Google, which will integrate Gemini models into parts of Apple’s AI platform. Together, these moves signal a shift away from a single-model strategy toward a layered AI architecture, where Apple selects the best tools for specific tasks while keeping control over privacy, user experience, and system-level integration.
Why “Runs on Anthropic”
When Gurman says Apple “runs on Anthropic,” he is pointing to the company’s internal reliance on Claude-based systems for certain development, evaluation, and model alignment workflows. Anthropic’s models are known for safety-focused design, structured reasoning, and conversational clarity. Those strengths align closely with what Apple has promised for its future assistants: reliable context, controlled responses, and a more human conversational tone.
Rather than positioning Anthropic as a competitor to Apple’s own models, the relationship appears to be complementary. Apple’s internal teams still design Foundation Models and on-device AI layers, but Anthropic provides a backbone for advanced reasoning and large-scale training support. This allows Apple to iterate faster while maintaining a high standard for output quality and safety.
The Role of Google Gemini
At the same time, Apple confirmed that it is partnering with Google to bring Gemini models into its AI ecosystem. These models will help power select Apple Intelligence features, including the next-generation version of Siri that Apple first previewed at WWDC 2024. The integration allows Apple to draw on Google’s scale in large language models while still processing sensitive data on-device or through its Private Cloud Compute infrastructure.
This multi-model approach is unusual for Apple, a company historically known for vertical integration. But AI is different. Training and maintaining competitive models requires enormous resources, and Apple appears to be choosing flexibility over isolation. By combining Anthropic for reasoning and safety workflows, Gemini for large-scale model capabilities, and Apple’s own models for privacy-focused tasks, the company is building a modular AI platform that can evolve quickly.
Two Versions of Siri
Gurman also reports that Apple is preparing two versions of Siri. One will run primarily on-device for everyday tasks, protecting user privacy and working even when offline. The other will rely on cloud-based intelligence, drawing from both Gemini and Anthropic-backed systems to handle complex, multi-step requests.
This dual-layer design reflects Apple’s broader philosophy: powerful intelligence without sacrificing trust. Simple commands stay local. More advanced interactions, such as planning, summarizing, or reasoning across multiple apps, can tap into cloud resources when needed. The user experience remains seamless, but the architecture underneath becomes far more sophisticated.
Management Changes and Strategic Realignment
Behind the scenes, Apple has also reorganized its AI leadership. Gurman notes that the company reshuffled internal teams to better align software, services, and machine learning groups. This restructuring is designed to remove bottlenecks that slowed development and to ensure that AI features move from research to products faster.
These changes suggest Apple is treating AI not as a single feature category, but as a system-wide capability. Every app, service, and device will increasingly rely on this new intelligence layer, from writing tools and image generation to voice control and contextual automation.
A New AI Identity for Apple
The statement that Apple “runs on Anthropic” is less about dependency and more about direction. Apple is no longer positioning itself as a closed AI lab competing in isolation. Instead, it is building an ecosystem of intelligence, where the best models are selected, integrated, and refined under Apple’s design and privacy standards.
For users, this means smarter interactions, more natural conversations with Siri, and tools that understand context rather than just commands. For Apple, it marks the beginning of a more open, adaptive approach to artificial intelligence—one that could define the next decade of its ecosystem.
