Report: Apple May Use Anthropic or OpenAI to Fix Siri

In a major reversal, Apple is looking into using Anthropic or OpenAI models to get conversational Siri to market more quickly.

That’s according to a report from the reliable Mark Gurman, who says that the change would require it to “sideline” its own in-house models. Apple has apparently spoken to both companies about using their models for Siri, and it is testing custom versions of each now on its Private Cloud Compute infrastructure. Apple currently uses its own models for all of Apple Intelligence, except for the web-based results that Siri delivers via OpenAI’s ChatGPT and some Writing Tools features.

Gurman says that Apple hasn’t decided yet which solution makes the most sense. But it’s continuing to improve its own in-house models, too, just in case. Either way, the goal is to close the gap with Google Gemini on Android and various third-party models. And in doing so, change its reputation as an AI wannabe.

Siri chief Mike Rockwell and Apple software engineering head Craig Federighi were apparently behind the decision to evaluate other models. And Gurman says that Anthropic Claude is currently the front-runner, based on what Apple needs for Siri. But Anthropic has apparently asked for multibillion dollar annual fees, and that’s not the Apple way.

Gurman notes that the external AI strategy is similar to what Samsung does with Galaxy AI, which is mostly based on Google’s Gemini, on its smartphones. And Amazon is also using Anthropic models for its Alexa+ service. He also reiterates that Apple has considered purchasing Perplexity AI, adding that it also held talks with Thinking Machines Lab, an AI startup founded by former OpenAI Chief Technology Officer Mira Murati.

Basically, everything is up for grabs at this point.