Apple vs. Android: The New Frontier of AI Privacy and Innovation

To go back to this article, go to My Profile and then click on View saved stories.

Apple's Privacy Measures Compared to Android's 'Hybrid AI'

During the Worldwide Developers Conference held on June 10, Apple made a significant, albeit delayed, foray into artificial intelligence with the launch of "Apple Intelligence." This announcement confirmed the long-standing speculation that Apple would collaborate with OpenAI to integrate ChatGPT into its iPhone lineup.

Elon Musk, a founding member of OpenAI, swiftly reacted on X, labeling AI tools powered by ChatGPT from Apple as "intrusive surveillance technology" and a "breach of security that cannot be tolerated."

"Musk stated that if Apple incorporates OpenAI into its operating system, then his companies will prohibit the use of Apple devices."

Amid growing concerns over AI privacy, Apple has introduced Apple Intelligence, promising a novel approach to safeguarding user data. The company is focusing on identifying which essential activities can be handled directly on the device.

In response to sophisticated demands, Apple has introduced a novel cloud-based framework named Private Cloud Compute (PCC), operating on its proprietary silicon servers. The firm asserts that this is a groundbreaking approach to safeguarding privacy during the emerging era of artificial intelligence.

Apple's chief of software engineering, Craig Federighi, describes the company's approach as setting a fresh benchmark for AI privacy standards. The question arises: do Apple's assertions hold water, and how does its approach stack up against the "hybrid AI" features found in devices like those in Samsung's Galaxy lineup?

AI Encounters E2E

Apple has introduced a revolutionary end-to-end AI framework and a secure cloud extension that acts as an extension of an iPhone user's device, according to Zak Doffman, CEO of Digital Barriers, a company focused on the real-time storage and analysis of surveillance footage. This innovation grants users enhanced data management capabilities.

In practical terms, this allows Apple to obscure where AI requests come from and block all parties, even the company that manufactures iPhones, from accessing your information. Doffman notes, "Theoretically, this is about as near to full encryption for cloud-based AI as possible."

Bruce Schneier, the head of security architecture at Inrupt, commends Apple for creating an "exceptionally robust privacy framework" for its artificial intelligence technology. He believes that Apple aims to ensure that utilizing AI, even on their cloud services, maintains a level of security equivalent to that of their smartphones. According to Schneier, despite the complexity of the system, Apple has managed to perform admirably.

Up to this point, there's nothing comparable. The "Hybrid AI" technology utilized in Samsung Galaxy devices that operate on Google Android, as well as in Google's Nano series, processes certain AI tasks on the device itself while using cloud services for more complex functions when needed.

Camden Woollven, who leads the AI division at GRC International Group, an IT governance company, emphasizes the goal of maximizing user privacy alongside delivering robust AI capabilities. "This could enable the deployment of advanced AI technologies even on moderately priced smartphones," he suggests.

However, the approach of combining on-device and cloud-based AI processing could introduce vulnerabilities, as certain information is transmitted to cloud servers, lacking the robust oversight provided by Apple's Private Compute Core (PCC). "In the case of hybrid AI, there's a necessity for some information to exit the device for external processing, which heightens the risk of it being intercepted or improperly used," explains Riccardo Ocleppo, the founder and director of the Open Institute of Technology, an organization that offers courses centered around technology.

Google and its associated hardware manufacturers maintain that privacy and security are central to their strategy for Android AI. Justin Choi, the Vice President and leader of the security team for mobile experience at Samsung Electronics, suggests that their hybrid AI system provides individuals with significant control over their personal information, ensuring robust privacy protections.

Choi explains that the characteristics managed in the cloud are safeguarded by servers that adhere to rigorous protocols. "The AI capabilities on our device add an additional layer of security because they execute functions directly on the device without depending on cloud servers. This means no data is stored on the device or transferred to the cloud," Choi states.

Google claims that its data centers employ stringent security protocols, encompassing physical safeguards, entry management, and data encryption. The company emphasizes that when handling AI inquiries through the cloud, the information remains within Google's secure data center framework, ensuring that your data is not transferred to external entities.

In a statement, Choi mentioned that the AI systems in Galaxy devices do not use data from the features on the device for training purposes. Samsung makes it evident which AI capabilities operate directly on the device by marking them with the Galaxy AI emblem. Additionally, the company incorporates a watermark to identify content that has been created using generative AI.

The company has now launched a novel feature known as Advanced Intelligence settings, allowing individuals to opt-out of cloud-based AI functions for enhanced security and privacy.

Google asserts its longstanding commitment to safeguarding users' data privacy, emphasizing that this commitment extends to its AI functionalities, both on-device and cloud-based. “For sensitive operations like screening phone calls, we employ models that operate on the device itself, ensuring data does not exit the device,” Suzanne Frey, Google's vice president of product trust, explained to WIRED.

Frey explains that Google's offerings depend on its cloud-based systems, ensuring that a user's data, especially confidential details intended for summarization, is never transferred to an external party for handling.

"Frey highlights that their dedication has always been towards developing artificial intelligence features which are inherently secure and designed with privacy in mind. These features, he emphasizes, are built in accordance with Google’s pioneering responsible AI standards, showcasing a commitment to trust and responsibility in AI advancements."

Apple Shifts the Discourse

Instead of just adopting the mixed method for handling data, specialists believe that Apple has altered the essence of the discussion regarding artificial intelligence. "The assumption was a focus on device-based, privacy-centric processing. However, Apple's stance was essentially that the specifics of AI activities or their location are secondary to the manner in which they're executed," explains Doffman. He suggests this perspective is set to establish the gold standard throughout the mobile AI industry.

Despite this, Apple's victory in the AI privacy war is not secure: The agreement with OpenAI, marking a rare move by Apple to allow an external company access to its iOS platform, might compromise its assertions of privacy.

Apple has countered Musk's allegations that its collaboration with OpenAI jeopardizes iPhone security, stating that there are inherent privacy safeguards for individuals using ChatGPT. According to the company, users will be prompted for their consent before any inquiry is transmitted to ChatGPT, IP addresses will be anonymized, and OpenAI is committed to not retaining queries. Nonetheless, users are reminded that ChatGPT's own data handling regulations remain in effect.

Collaborating with another firm represents an “unusual strategy” for Apple, according to Jake Moore, a worldwide cybersecurity consultant at ESET. He emphasizes that such a choice “would have been carefully considered,” despite the current uncertainty surrounding the privacy outcomes. Moore acknowledges the possibility that “personal information might be gathered by both parties and possibly examined by OpenAI.”

Schneier points out that in certain cases, individuals have to abide by OpenAI's regulations. He notes that although Apple removes personal details before forwarding queries to OpenAI, a significant amount of queries still contain information that could reveal someone's identity.

The partnership between Apple and OpenAI could fundamentally transform how responsibility is managed across the field of AI, according to Andy Pardoe, an AI professor and the founder of Wisdom Works Group, a technology consultancy. "Such collaborations spread out the accountability among several participants, influencing the way in which credit for achievements and blame for possible shortcomings are apportioned within their joint operations."

Integrating AI technology into the operating systems of major tech companies introduces potential security vulnerabilities. According to Doffman, although Apple Intelligence might seem remarkable in concept, it brings forth an unprecedented scope for cyber threats. He explains that this technology amalgamates numerous security advancements for a novel application, positioning it as a potential transformative development. However, Doffman notes that upon its deployment, Apple will inevitably have to address emerging challenges.

This is why Apple and Google are motivating cybersecurity experts to uncover vulnerabilities in their artificial intelligence technologies. Google’s Secure AI Framework involves the cybersecurity firm Mandiant in evaluating the robustness of AI models.

A representative from Apple shared references to its PCC security document and the framework paper that outlines its conscientious strategy towards AI creation. Apple refrained from giving a direct response to WIRED's additional inquiries about privacy issues linked to its collaboration with OpenAI or the possible security threats involved.

Apple has introduced a concept known as "verifiable transparency" for its Private Cloud Compute (PCC). Through a recent blog post, Apple explained its intention to allow security and privacy experts to scrutinize the PCC software, verify its operations, and assist in pinpointing potential problems, much in the same way they are able to with Apple hardware. To facilitate this, Apple is providing access to software images of every production version of PCC for security researchers to examine.

The upcoming iOS 18 update is set to feature Apple Intelligence, which will be combined with ChatGPT functionalities. This comprehensive set of capabilities is expected to debut on the iPhone 16 later this year. Users will have the option to disable this feature, however, it's advisable to ponder the potential privacy and security risks before utilizing AI technologies on any gadget.

In the decision-making process between iOS and Android AI, the core issue is about trust, according to Pardoe. He suggests reviewing each operating system's balance of privacy options, how they manage data, and their openness. Pardoe emphasizes, however, that Apple's emphasis on privacy continues to stand out for individuals who place the utmost importance on protecting their data.

Recommended for You …

Delivered to your email: Fast Forward by Will Knight delves into the progress of artificial intelligence.

Delving into the largest undercover operation ever conducted by the FBI

The WIRED Artificial Intelligence Elections Initiative: Monitoring over 60 international electoral events

Ecuador finds itself completely helpless amid a severe drought.

Be confident: Here are the top mattresses available for purchase on the internet

Additional Coverage from WIRED

Insights and Manuals

© 2024 Condé Nast. All rights reserved. WIRED may receive a share of revenue from items sold via our website, as a result of our Affiliate Agreements with retail partners. Content from this website is not to be copied, distributed, transmitted, stored, or utilized in any form without explicit written consent from Condé Nast. Advertising Options.

Choose a global website