Siri

• 1. Introduction

  ›

  Siri was not born inside Apple. The voice assistant that would come to inhabit hundreds of millions of iPhones started life at SRI International, the Stanford Research Institute, as an offshoot of a DARPA-funded project called CALO. The woman whose voice became Siri's American identity, Susan Bennett, recorded her part in July 2005 without any idea what she was recording it for. By the time Siri arrived on the iPhone 4S on the 4th of October 2011, it was already years old in spirit, having passed through a defense research lab, a startup, and an Apple acquisition. How did a DARPA spinout become the voice of a trillion-dollar company? And why, for all its early promise, would Siri spend much of the following decade playing catch-up to rivals it had helped inspire?

• 2. Origins and Acquisition

  ›

  Dag Kittlaus, Tom Gruber, and Adam Cheyer co-founded Siri as a company built on research from SRI International's Artificial Intelligence Center. The name Kittlaus chose came from a former co-worker in Norway; Siri is a shortened form of Sigrid, rooted in Old Norse elements meaning "victory" and "beautiful."

  The underlying speech recognition was not Apple's own work. It came from Nuance Communications, a speech technology company. Neither Apple nor Nuance acknowledged this arrangement for years, until Nuance's CEO Paul Ricci confirmed it at a technology conference in 2013. The system Nuance supplied used convolutional neural networks and long short-term memory, techniques drawn from machine learning research.

  The first Siri prototype ran on something called the Active platform, a joint project between SRI International's Artificial Intelligence Center and the Vrai Group at Ecole Polytechnique Federale de Lausanne. Didier Guzzoni, whose doctoral thesis focused on the Active platform, joined Siri as its chief scientist. Apple had imagined a digital personal assistant as far back as 1987, when a concept video called Knowledge Navigator sketched out the idea. Apple acquired the Siri company in April 2010 under the direction of Steve Jobs, two years before the public had ever heard the name.

• 3. The Voices Behind the Voice

  ›

  Susan Bennett recorded the American voice of Siri in July 2005, unaware the material would power a global product. For years, the identity of Siri's voice was a subject of speculation. A report from The Verge in September 2013 suggested Allison Dufty was the voice, but Dufty flatly denied it on her own website, writing that she was "absolutely, positively not the voice of Siri." Facing growing pressure, Bennett revealed her role in October of that year, and audio forensics expert Ed Primeau confirmed her claim. Apple has never officially acknowledged it.

  The original British male voice came from Jon Briggs, a former technology journalist who had spent twelve years narrating the hit BBC quiz show The Weakest Link. Briggs learned he was Siri's voice by watching television. He later said the recordings were done "five or six years ago" without any indication of their intended use. In an interview with The Guardian alongside his fellow voice actors, Briggs explained that the original voice work was recorded for a US company called Scansoft, which was subsequently bought by Nuance, and that Apple simply licensed it.

  The original Australian voice belonged to Karen Jacobsen, a voice-over artist known in Australia as the GPS girl. When Apple rebuilt Siri's voice for iOS 11, it auditioned hundreds of candidates and recorded several hours of speech to construct a new text-to-speech voice using deep learning. In February 2022, Apple added Quinn to the iOS 15.4 developer release, its first gender-neutral voice and fifth user option.

• 4. Mixed Reviews and Early Criticism

  ›

  Siri's beta debut on the iPhone 4S drew genuinely divided reactions. MG Siegler of TechCrunch called it "great" and credited its contextual understanding, while noting there was "no API that any developer can use." Writing for The New York Times, David Pogue praised Siri's ability to understand context. Jacqui Cheng of Ars Technica acknowledged Siri could "decipher our casual language" and return specific results, but documented specific failures. One example Cheng highlighted: the instruction "Send a text to Jason, Clint, Sam, and Lee saying we're having dinner at Silver Cloud" caused Siri to send a message only to Jason, containing the names of the others as part of the message text.

  The launch was overshadowed almost immediately. Steve Jobs died one day after Siri was introduced. Google's executive chairman Eric Schmidt conceded that Siri could pose a competitive threat to the company's core search business, a recognition that Apple had touched something significant.

  Criticism expanded beyond technical shortcomings. Pro-abortion rights organizations including the ACLU and NARAL Pro-Choice America found that Siri could not locate birth control or abortion providers, sometimes directing users to crisis pregnancy centers. Apple's spokeswoman Natalie Kerris told The New York Times that "these are not intentional omissions." By January 2016, Fast Company reported that Siri had begun confusing "abortion" with "adoption," a development that health experts said had "gotten worse," even as the situation had partially improved with Siri offering a more comprehensive list of Planned Parenthood facilities.

  In March 2012, Frank M. Fazio filed a class action lawsuit in California on behalf of iPhone 4S buyers who felt misled by Apple's Siri commercials, calling the phone "a more expensive iPhone 4" if Siri failed as advertised. On the 22nd of July 2013, U.S. District Judge Claudia Wilken dismissed the suit, ruling that plaintiffs had not documented enough misrepresentations for the case to proceed, though she left open the possibility of amendment.

• 5. The Innovation Gap

  ›

  By the mid-2010s, a pattern of reporting had taken hold: Siri was falling behind. In June 2016, The Verge's Sean O'Kane wrote that Siri's biggest flaws remained its inability to reliably understand users, and that "when it gets it right, the results are often clunky." Walt Mossberg, also writing for The Verge, observed that after Apple acquired Siri, the company had treated it "as a backwater," limiting it to a slowly expanding list of simple tasks while its founders eventually left Apple to build a competing service called Viv.

  The Wall Street Journal published an extensive report in June 2017 on Siri's stagnation relative to Amazon's Alexa, Microsoft's Cortana, and Google's Assistant. The Journal reported that Apple workers' anxiety levels "went up a notch" when Amazon announced Alexa, and wrote that "Apple is playing catch-up in a product category it invented." The primary causes the report identified were Apple's strict approach to user privacy, including the practice of randomly tagging and discarding Siri search data after six months, in contrast to Google and Amazon retaining data until users actively removed it, and internal executive power struggles.

  Apple's Eddy Cue responded by noting that the company "often uses generic data rather than user data to train its systems" and has the ability to improve Siri for individual users using information kept on their own devices. In October 2016, Bloomberg had reported a plan to unify Apple's cloud-based teams onto a single campus and reorganize cloud computing resources to improve Siri's query processing. The structural problems, however, proved persistent.

• 6. Privacy Controversy

  ›

  In July 2019, former Apple contractor Thomas le Bonniec, then anonymous, disclosed that Siri regularly recorded user conversations, often triggered unintentionally, and sent those recordings to Apple contractors who graded Siri's responses. Contractors heard private conversations between doctors and patients, business and drug deals, and couples having sex. Apple had not disclosed this practice and had provided no opt-in or opt-out mechanism.

  Apple apologized in August 2019, suspended the grading program, and announced it would not listen to Siri recordings without user permission. iOS 13.2, released in October 2019, introduced controls allowing users to opt out of grading and to delete stored voice recordings. The program became opt-in.

  Le Bonniec revealed his identity in May 2020 and sent a letter to European data protection regulators urging an investigation into Apple's "past and present" use of Siri recordings, arguing the company had never faced consequences for the years-long program. In December 2024, Apple agreed to a $95 million class-action settlement to compensate users of Siri-enabled devices from the previous ten years. As part of the settlement, Apple must confirm the deletion of Siri recordings collected before 2019 and issue fresh guidance on data collection and user participation.

• 7. Expansion and Third-Party Access

  ›

  When Siri launched as a standalone app for iOS in February 2010, its developers planned to release it for Android and BlackBerry as well. Two months later, Apple acquired the company and those plans evaporated. On the 4th of October 2011, the iPhone 4S introduced Siri as a beta feature, and Apple pulled the standalone app from the App Store the same day.

  Third-party developer Steven Troughton-Smith managed to port Siri to the older iPhone 4, though without the ability to reach Apple's servers. Working with an anonymous collaborator known as "Chpwn," Troughton-Smith subsequently completed a full Siri hack enabling its functions on iPhone 4 and iPod Touch. Other developers created and distributed legal ports to any device running iOS 5, using a proxy server to handle Apple server communication.

  Apple expanded official support incrementally. iPad gained support in June 2012. iPod Touch followed in September 2012. Apple TV support and the standalone Siri Remote came in September 2015. Mac and AirPods joined in September 2016, and HomePod gained support in February 2018.

  The opening to third-party developers came with iOS 10 in June 2016, when Apple introduced a dedicated API allowing Siri to interact with messaging apps, payment apps, ride-sharing apps, and internet calling apps. iOS 11 expanded that to task management and added follow-up question support, language translation, and a privacy-focused on-device learning technique.

  At the 2021 Worldwide Developers Conference, Apple announced Siri voice integration for third-party hardware, routing requests through a HomePod or HomePod Mini on the same wireless network. The Ecobee SmartThermostat with Voice Control became the first third-party device with built-in Siri in October 2021. In 2024, Denon added Siri control to select soundbars and smart speakers.

• 8. Apple Intelligence and What Comes Next

  ›

  iOS 18, iPadOS 18, and macOS Sequoia brought large-language-model capabilities to Siri under the Apple Intelligence banner, including ChatGPT integration and a new interface allowing users to interact via text by double-tapping the home bar. Apple described the upgraded Siri as capable of using the context of device activity to make conversations more natural, delivering personalized results from on-device information, and summarizing messages across apps including Discord and Slack. Users described the summarization feature as sometimes helpful and sometimes inappropriate.

  In 2025, Apple announced that a broader Siri overhaul intended to enable deeper personalization would be delayed due to technical challenges. The following year, Apple announced that it had selected Google Gemini to power AI features including Siri. The $95 million privacy settlement reached in December 2024, which required Apple to confirm deletion of recordings gathered before the opt-in change, will stand as one of the most concrete consequences of the gap between what Siri collected and what users were told.