According to Bloomberg, Job’s Mob is considering using AI models from Anthropic or OpenAI in a desperate bid to make Siri look clever. While the company has been telling the world+plus dog that its own Apple Intelligence produces will be the best thing since sliced bread it has yet to get a decent product in the shops.
Behind the scenes, the black shirts in Cupertino have reportedly chatted with both firms about training large language models to run on Apple’s cloud gear. No decisions have been made yet, and a Siri version still running Apple’s wonky model is apparently being worked on too.
A new AI-powered Siri was supposed to be part of iOS 18, iPadOS 18, and macOS 15, but the usual delays have kicked that down the road until 2026. In the meantime, Apple has already quietly been using OpenAI to answer search queries and generate content, though only if users enable it.
Anthropic or OpenAI integration could give Apple the AI cred it sorely lacks. Samsung has been using Google’s Gemini models to power its Galaxy AI, and Apple is clearly struggling to keep up. There were even whispers it considered buying Perplexity outright.
After the iPhone 16 was hyped as being “built for Apple Intelligence,” things have unravelled. Siri and the Apple Intelligence team were handed off from AI chief John Giannandrea to software bloke Craig Federighi and Siri’s new head, Mike Rockwell. Rockwell previously oversaw the Vision Pro disaster.
Bloomberg claims morale in the flying saucer is circling the drain inside Apple’s AI teams, as outside help threatens their relevance just when rival tech firms are splashing the cash for top AI talent.
The rot isn’t just in Siri. Losing it meant Giannandrea saw Apple’s robotics division shuffled out of reach. Meanwhile, a long-rumoured smart home Franken-device, a cross between a HomePod and an iPad, has been put on ice too.
At WWDC, Job’s Mob still insisted it’s sticking with its own models for things like Genmoji and Writing Tools, and plans to let developers build features using these kits. However, those on-device models are slower and less capable than the competition’s cloud-based offerings.