Apple has lost its fourth AI researcher in a month to Meta Platforms, marking the latest setback to the iPhone maker’s artificial intelligence efforts.

Bowen Zhang, a key multimodal AI researcher at Apple, left the company on Friday and is set to join Meta’s recently formed superintelligence team, according to people familiar with the matter. Zhang was part of the Apple foundation models group, or AFM, which built the core technology behind the company’s AI platform.

Meta previously lured away the leader of the team, Ruoming Pang, with a compensation package valued at more than US$200-million. Two other researchers from that group — Tom Gunter and Mark Lee — also recently joined Meta. AFM is made up of several dozen engineers and researchers across Cupertino, California and New York.

In response to the job offers from Meta and others, Apple has been marginally increasing the pay of its AFM staffers, whether or not they’ve threatened to leave, said the people, who asked not to be identified because the moves are private. Still, the pay levels pale in comparison with those of rivals.

Spokespeople for Apple and Meta declined to comment.

Apple shares declined as much as 1.5% to $210.82, reaching a session low in New York trading. The stock was already down 15% this year through to Monday’s close.

The departures have thrown Apple’s models team into flux. Pang played a central role in defining the department’s road map and research direction, and multiple people within AFM now say its future is unclear. Additional engineers are actively interviewing for jobs jobs elsewhere, according to the people. Another team member — Floris Weers — left for a start-up in recent weeks.

Apple Intelligence

The AFM team is critical to Apple’s broader AI strategy. The group’s work underpins the Apple Intelligence platform, which launched last year. But now the company is considering a shift towards using more third-party models.

Some Apple executives see its homegrown models as a stumbling block to catching up with AI rivals, the people said. And the uncertainty over whether to outsource the technology has hurt morale at the company and helped fuel the attrition.

Meta, meanwhile, is aggressively staffing up. The Facebook owner has offered gigantic compensation packages to AI talent across Silicon Valley, luring away workers from Apple, OpenAI and Anthropic.

Read: Meta to build Manhattan-scale, multi-gigawatt data centres

In recent months, Apple started considering a move away from AFM models for a new version of its Siri voice assistant. The work includes powering Siri with OpenAI’s ChatGPT or Anthropic PBC’s Claude models.

The company is simultaneously working on a competing version based on new AFM models. While a final decision hasn’t been made, Apple’s exploration of outside options has triggered unease within AFM.

Internally, executives have sought to reassure the team members, saying their work remains important to Apple’s AI strategy. They have told engineers that the company is committed to in-house model development, part of a broader desire to own critical underlying technologies, as the company has done with chips in recent years.

But Apple’s own policies have made it harder for its AI team to keep up with competitors. The company has a longstanding commitment to privacy and generally favours handling AI tasks on devices — rather than in the cloud — so the data doesn’t have to be processed somewhere out of users’ control. That approach limits AI capabilities because phones aren’t as powerful as data centres.

Apple Intelligence mostly relies on an on-device model with three billion parameters, a measure of complexity and learning capacity. Competitors, in contrast, offer cloud-based systems with a trillion-plus parameters. Apple does have its own cloud model, but that’s in the range of 150 billion parameters.