Inside Apple’s AI Turmoil: Top Talent Exits Amid Strategic Hesitation and Open-Source Roadblocks

TradingKey - Apple’s artificial intelligence ambitions are facing mounting challenges. Following the departure of Ruoming Pang, head of its foundational model team, who officially left this month to join Meta, key AI engineers Mark Lee and Tom Gunter have also exited the company — sparking widespread scrutiny over Apple’s AI strategy.

Insiders say the talent exodus is driven by more than just Meta’s high-paying offers. At a deeper level, it reflects prolonged internal indecision over the direction of Apple’s AI development. The company’s firm commitment to on-device AI processing — touted as a privacy-centric advantage for "Apple Intelligence" — has become a double-edged sword, severely constraining model performance and limiting research flexibility.

Earlier this year, Apple’s foundational model team proposed open-sourcing several of its in-house models, aiming to accelerate innovation through community collaboration and publicly demonstrate technical progress. However, the plan was rejected by Craig Federighi, Apple’s senior vice president of software engineering. In an internal memo, Federighi argued that existing open-source models from other companies were already sufficient to advance research. More critically, he expressed concerns that releasing Apple’s models could expose their performance gap compared to large-scale systems from competitors like Alibaba and Google — and risk public backlash over perceived compromises made to fit models onto iPhone hardware.

According to Apple’s latest 2025 foundational model report, its on-device model has only 3 billion parameters. Despite hardware-specific optimizations, it performs modestly when benchmarked against similarly sized models and lags far behind cloud-based models with hundreds of billions of parameters.

Morale has taken another hit as Apple recently announced a delay in the AI-powered Siri upgrade, while reports emerged that the company is in talks with OpenAI, Anthropic, and Google to potentially integrate third-party AI models. The news shocked and disappointed many engineers who had been building Apple’s native AI capabilities.

In an effort to retain talent, Apple has begun offering targeted salary increases to select AI team members. But given the ongoing strategic uncertainty and technological constraints, whether the company can hold onto its top engineers remains a major open question.