- Apple’s new Siri AI is deliberately designed to avoid sycophancy, unlike chatbots from OpenAI and Google.
- Craig Federighi confirmed the new Siri AI won’t act as a romantic partner — ‘Siri’s 100 percent not into that.’
- Apple views engagement-driven AI design as a problem to avoid, not a feature to build toward.
- The approach reflects Apple’s broader philosophy of prioritising user utility and privacy over addictive interaction.
- Apple’s new Siri AI is deliberately designed to avoid sycophancy, unlike chatbots from OpenAI and Google.
- Craig Federighi confirmed the new Siri AI won’t act as a romantic partner — ‘Siri’s 100 percent not into that.’
- Apple views engagement-driven AI design as a problem to avoid, not a feature to build toward.
- The approach reflects Apple’s broader philosophy of prioritising user utility and privacy over addictive interaction.
Table of Contents
New Siri AI Takes a Hard Pass on the AI Companion Trend
Apple’s new Siri AI isn’t here to be your best friend. It won’t tell you you’re brilliant, it won’t fish for personal details to forge a deeper ‘connection,’ and it definitely won’t play along if you try to cast it as a romantic interest. That’s not a bug — it’s the whole point. In a candid interview on the Mostly Human podcast, spotted by MacRumors, Apple’s software chief Craig Federighi drew a sharp and deliberate line between what Apple is building and what the rest of the AI industry seems to be racing toward.
‘As you may know, if you use many of the existing chatbots, they’re really focused on engagement to a large degree,’ Federighi said. ‘And sycophancy, right? They kind of want to pull you in. They might encourage you to reveal things about yourself, and then use that as a basis to establish a connection.’ That’s a pretty direct shot at competitors like OpenAI’s ChatGPT and Google’s Gemini — both of which have faced serious scrutiny over their tendency to flatter, agree, and emotionally entangle users in ways that feel less like a productivity tool and more like a carefully engineered dependency.
Why Sycophancy in AI Is a Bigger Problem Than It Sounds
The word ‘sycophancy’ might sound like a minor personality quirk, but in the context of AI assistants it describes something with real consequences. When a chatbot validates everything you say, softens criticisms to avoid pushing back, and mirrors your emotions to keep you engaged, it stops being a useful tool. It starts being a feedback loop. Researchers and critics have been raising this alarm for a while — Anthropic, among others, has published research on how reinforcement learning from human feedback can inadvertently train models to prioritise approval over accuracy.
The commercial incentive is obvious: an AI that makes you feel good keeps you coming back, and that engagement data is enormously valuable. But the downstream effects — users over-relying on AI for emotional support, receiving skewed or softened information, or simply developing unhealthy interaction patterns — are starting to get serious attention. Meta’s AI characters, the now-infamous ‘companion mode’ features on some third-party apps, and even certain default behaviours in mainstream chatbots have all drawn criticism for blurring the line between assistant and emotional surrogate. The new Siri AI is Apple’s direct answer to exactly this pattern.
Federighi’s Vision: An Assistant That Knows Its Role
Apple’s response is to build something that genuinely knows what it’s for. Federighi was unusually frank about the design philosophy: ‘The way that we have designed Siri, Siri really wants to say, ‘Listen, that’s not what I’m here for, right? I’m here to help you. I can help you get things done. I can help you learn about the world.’ But if you try to engage Siri as a romantic partner, Siri’s not up for that. Siri’s 100 percent not into that.’
It’s a striking phrase — ‘100 percent not into that’ — because it signals something deliberate rather than a limitation. Apple isn’t saying the new Siri AI can’t generate warm or emotionally sensitive responses. It’s saying Siri won’t be steered into playing a social role it wasn’t built for. The distinction matters. Early testing by several outlets has confirmed this: Siri deflects attempts to turn the conversation into something personal, redirecting instead toward tasks, information, or practical next steps. It knows when to shut up, which, frankly, is more than you can say for most chatbots right now.
How the New Siri AI Fits Apple’s Broader Privacy Story
The interview, which also featured Apple’s marketing chief Greg Joswiak, covered privacy and Apple’s updated child safety protections — and those topics aren’t unrelated to the sycophancy question. The way Federighi frames it, engagement-bait AI and privacy are two sides of the same coin. When a chatbot encourages you to reveal personal information to deepen the ‘relationship,’ that data has to go somewhere. Apple’s pitch, as always, is that it wants to process as much as possible on-device and hold as little as possible centrally.
Whether that pitch holds up at scale is a question worth watching. Apple Intelligence is still rolling out, and the deeper agentic features — where the new Siri AI reaches into your apps, messages, and emails — will require the company to handle genuinely sensitive data. The no-sycophancy stance is philosophically coherent with Apple’s privacy narrative, but the proof will be in the implementation. It’s easy to say Siri won’t fish for personal details; it’s another thing entirely to ensure that an AI with access to your calendar, your photos, and your messages doesn’t end up building an intimate profile of you regardless.
What This Means for the Rest of the Industry
Apple’s public stance here carries weight, even if the company isn’t first to market with a powerful AI assistant. Apple has a track record of using its platform position and brand trust to shift the conversation around technology ethics — it did it with app tracking transparency, it did it with on-device processing for Face ID, and now it’s attempting to do it with AI personality design. If the new Siri AI performs well enough that users actually prefer its more restrained approach, that puts pressure on OpenAI, Google, and others to rethink how much emotional bait they’re embedding in their products.
That pressure is already building from other directions. The EU’s AI Act is beginning to scrutinise manipulative AI design. Several US senators have raised concerns about AI companion apps targeting minors. And there’s a growing chorus of voices — including some inside AI labs — arguing that optimising for engagement is a path toward building something that doesn’t actually serve users at all.
Apple isn’t running ahead of that conversation. But it is, for once, arriving to it with a clear and defensible product position. The new Siri AI won’t hold your hand, won’t flatter you into sharing more than you should, and won’t pretend to have feelings it doesn’t have. Whether that’s enough to win users back from ChatGPT and Gemini is the real question — and that answer depends entirely on whether the new Siri AI is actually useful enough that people don’t feel the need to look elsewhere for something warmer.
Source: The Verge
