Ready for a Siri AI agent? Apple under pressure to upgrade its assistant faster

ZDNET’s cardinal takeaways

• Bloomberg’s Mark Gurman reports Apple whitethorn propulsion harder connected its AI strategy.
• That would mean a much agentic Siri.
• Apple could suffer crushed to AI-first rivals.

Apple’s precocious appointed CEO, John Ternus, is allegedly pushing the institution internally to ramp up its AI strategy — including making Siri an AI agent.

Also: My 5 favourite Siri AI ‘onscreen awareness’ tricks

According to Bloomberg’s Mark Gurman, the CEO is publically pleased with the Siri AI launch, but privately encouraging Craig Federighi, Apple’s SVP of bundle engineering, to determination faster.

Siri AI and Apple’s conflict

The revamped Siri AI is present a overmuch amended dependable adjunct with immoderate adjuvant features. Gurman noted helium thinks Siri AI’s quality to automatically fix your anemic passwords is already somewhat agentic behavior, though AI-generated passwords have their ain issues.

But Siri AI isn’t yet connected par with progressively autonomous agents from AI-first companies that purpose to proactively signifier your tasks, usage your computer, and constitute precocious code. Still, those rival agents present a patchwork of capabilities: portion some, like OpenClaw and Anthropic’s Claude Code, are highly functional, others, similar Google’s Gemini Spark, aren’t rather determination yet.

Also: I tried replacing Claude Code with a escaped section AI stack

Regardless, the contented present is time. Despite developing almighty hardware perfectly suited for moving section AI models, Apple has remained slower than competitors connected the AI bundle front. Competitors person demonstrated they tin marque promised cause features a reality, immoderate quicker than others.

Apple’s slower gait successful AI has, successful the eyes of many, really benefited its estimation amid a rising tide of nationalist backlash against AI. Some users are frustrated by the software bloat from mediocre oregon unnecessary AI features, particularly arsenic hardware prices skyrocket to accommodate them during a representation shortage. Users besides instrumentality comfortableness successful Apple’s privacy-first attack to processing AI, which has helped it basal retired against different AI products.

Meta’s Muse cause mightiness beryllium an utmost illustration of data-hungry AI agents, but its privacy trade-offs aren’t altogether absent from different idiosyncratic agents. Gemini relies connected much of your data to beryllium the astir effectual mentation of itself, pitting privacy risk against usability.

Also: Meta Muse is the worst AI cause for privacy

On the different hand, the information that Apple hyped Siri AI for 2 years lone to merchandise an improved dependable assistant (rather than a groundbreaking AI product) whitethorn mean users are acceptable for much precocious features.

In addition, AI companies are working connected their ain hardware. “If they tin harvester that bundle with AI-native devices designed from the crushed up astir assistants that see, retrieve and act, Apple’s longstanding advantages could statesman to substance less,” Gurman wrote.

What an agentic Siri AI could mean

An agentic Siri would astir apt physique connected what users already similar astir the caller version.

The on-screen awareness feature, for example, is presently constricted to what’s disposable connected your surface successful that moment, meaning lone the portion of a agelong email that you’ve actively scrolled to. An agentic Siri could ideally scroll each the mode done an email connected its ain and aboveground the insights you’re looking for.

Also: The iPhone made Apple a elephantine – but the Mac could assistance it triumph the AI era

But successful the abbreviated term, Ternus’ propulsion volition apt conscionable present connected existing promises. Gurman besides pointed retired that Siri AI inactive lacks precocious in-app actions, which Apple initially announced successful 2024 and would fto third-party apps make plug-ins for and pathways done Siri.

“This functionality would fto Siri pat into some in-house and third-party applications with much precision,” Gurman wrote. “This past week, arsenic portion of the first beta mentation of iOS 27.3, Apple yet began talking astir the diagnostic again.”

If Siri AI were a competitory agent, galore Apple users who spot the company’s ecosystem mightiness determination distant from installing apps similar ChatGPT oregon Claude connected their devices. That’s inactive improbable arsenic of present fixed the wide spread betwixt Apple’s AI and user products from OpenAI and Anthropic, but not disconnected the array arsenic nationalist AI backlash and privateness concerns mount.

(Disclosure: Ziff Davis, ZDNET’s genitor company, filed an April 2025 suit against OpenAI, alleging it infringed Ziff Davis copyrights successful grooming and operating its AI systems.)

Plus, if we’re talking astir agents specifically, newcomers similar OpenAI’s Dot and Anthropic’s Duet aren’t competitory themselves yet: ZDNET’s AI vetter-in-residence David Gewirtz found some to beryllium “a small odd” and not “quite acceptable for premier time”.

Also: I tested OpenAI Dot and Anthropic Duet for weeks – and each I got was 1 adjuvant notification

ChatGPT’s Computer History feature — which creates a timeline of your app and website activity, and is lone disposable connected the Mac — raised privateness concerns erstwhile it launched successful August. While the diagnostic is opt-in and tin beryllium turned disconnected astatine immoderate time, OpenAI said it does not encrypt delicate accusation the diagnostic sees, which leaves it susceptible to hackers.

If Apple pushes guardant with AI, it could plan much unafraid versions of features similar these in-house, relying connected its well-known and trusted committedness to privateness and to protecting idiosyncratic data.

For users who admit the subtlety of Siri AI, though, thing akin to ChatGPT oregon Claude connected their devices mightiness not land, unless each diagnostic is opt-in.

Radhika Rajkumar

Senior Editor

Radhika Rajkumar is simply a elder exertion astatine ZDNET based successful New York City. She covers AI, specializing successful safety, privateness and security, policy, education, and synthetic media. She besides leads ZDNET's newsletter strategy. Radhika holds a Masters successful Creative Publishing and Critical Journalism from The New School. See afloat bio