Foldable iPhone Duo Pre-Orders Open: AI Agents Race to Buy and the AI Privacy Crisis

Apple's first foldable phone, the iPhone Duo, will open for pre-orders tomorrow, but according to The Verge's Vergecast podcast, the "pre-pre-order" has already begun — Apple is letting users complete everything in advance, from choosing their model to entering payment information, leaving only the final purchase button to be pressed tomorrow morning. For ordinary consumers, the real competition may not be scalpers, but AI agents doing the same thing.

iPhone Duo's "Pre-Pre-Order" and the AI Buying Rush

According to The Verge, Apple is adopting a more aggressive sales cadence: users can complete the entire configuration process except for payment before the official pre-order. This means that when sales open tomorrow morning, conversion speed will be dramatically higher.

The real wildcard is AI agents. When users can delegate AI agents to make purchases on their behalf, ordinary consumers placing orders manually have virtually no chance against machine speed. The Vergecast's commentary was tinged with humor: "Good luck beating the AI agents who also want to buy the phone." This actually points to a broader trend — once shopping gateways open to AI agents, retail platforms' inventory allocation rules, anti-bot mechanisms, and even the definition of a "fair right to purchase" all need to be rewritten.

What is AI agent buying?

An AI agent refers to a software program capable of autonomously completing multi-step tasks on a user's behalf. It doesn't just passively respond to commands — it can plan workflows, call APIs, and execute continuously. In the context of flash buying, users can pre-authorize an agent to submit orders at a specified time, automatically fill in payment information, bypass page-loading delays, and even monitor multiple inventory channels simultaneously.

What makes this threatening to ordinary consumers is the sheer difference in speed and concurrency. A human typically needs several seconds from seeing the page to clicking to order, while an agent can complete the same action in milliseconds and run thousands of instances simultaneously via scripts.

This raises platform governance issues: traditional anti-bot mechanisms (CAPTCHAs, behavioral analysis, purchase limits) were largely designed against "scalper scripts." But when an AI agent appears as a "legitimate user-authorized assistant," it's hard for platforms to determine whether it violates the rules. In the future, we may need verifiable agent identities, rate limiting by user rather than by request, and even a redefinition of "who is eligible to buy."

Apple and OpenAI's Trade Secret Lawsuit

The lawsuit between Apple and OpenAI, filed in July, has reached a new turning point. OpenAI has promised not to use hardware secrets it obtained from some former Apple employees. The timing of this commitment is rather delicate — it came shortly after a U.S. district judge indicated she was likely to approve an injunction forcing OpenAI to do so.

The accusations are mutual: OpenAI simultaneously accuses a former employee of sharing confidential documents with Apple. At the heart of this tug-of-war is the intellectual property boundary in hardware talent mobility. As both Apple and OpenAI advance their own hardware product lines, such disputes are unlikely to be isolated cases — they will become routine friction in the AI hardware race.

Trade secret litigation and the boundaries of "non-compete"

Unlike patents, trade secrets require no public registration and protect undisclosed technical information, design documents, supply chain details, and more. The U.S. Defend Trade Secrets Act (DTSA) allows rights holders to seek an injunction when they can prove that "misappropriation is likely to continue," which is the basis for the judge potentially granting an injunction in this case.

Hardware talent mobility in Silicon Valley has historically been accompanied by such disputes, because the line between an engineer's personal experience and company secrets is extremely hard to draw. What an employee takes when changing jobs may be "general knowledge in their head," or it may be specific blueprints or supplier lists. The legal standard for distinguishing is usually whether the information was subject to confidentiality measures and whether it has independent economic value.

OpenAI's proactive commitment not to use the relevant secrets is essentially a "voluntary compliance" strategy to avoid an injunction taking effect, reducing legal risk while preserving room for future negotiations. As the AI hardware race heats up, such lawsuits are likely to evolve from isolated cases into an industry norm.

The Privacy Predicament in the Age of Always-Listening Devices

The Vergecast raised a sharp observation: as always-listening wearables proliferate, "we'll all soon have no secrets left." This isn't alarmism — it's dictated by the technical architecture. Devices must continuously process audio either locally or in the cloud, naturally expanding the data exposure surface.

The contradiction here is that users willingly accept an always-on microphone for convenience, but most don't know how their data is processed, how long it's stored, or who can access it. When an AI agent both helps you shop and continuously listens to you, the line between convenience and surveillance becomes extremely blurred.

Why always-listening device architecture struggles to keep secrets

To achieve "always ready to wake," always-listening devices typically need to run a low-power wake-word detection model locally, then upload audio to the cloud for further processing after waking. This means the microphone can hardly ever truly be off — it can only be in a state of "local buffering, on-demand uploading."

The risks come mainly from three levels: first, locally cached audio clips, if unencrypted, can be extracted through physical access or system vulnerabilities; second, stored conversation records in the cloud, if poorly managed, may be accessed by insiders or attackers; third, device manufacturers often retain data under the pretext of "improving services," leading to storage periods far longer than users expect.

More subtly, users typically see privacy terms only once during initial setup and afterward have little awareness of whether the device is collecting. At the regulatory level, most regions currently require "notice and consent," but the continuous and covert nature of always-listening devices significantly undermines the practical effectiveness of this principle.

AI Pet Eleco Data Breach: Millions of Conversations Unprotected

According to The Verge's Sean Hollister, researchers discovered security vulnerabilities in an AI pet toy called Eleco. Manufactured by Energize Labs, the toy transcribed millions of conversations with no protection measures whatsoever.

Even more disturbing is what followed: the company used these conversations to profile customers. One customer was described as "difficult to trust family members," and the company wrote at the end of its report that "after this experience, this customer's trust issues will only continue to worsen." This isn't an abstract data leak — it's turning the most intimate psychological weaknesses into corporate files. For any company selling AI hardware to children or families, Eleco is a direct cautionary tale.

Why "profiling" in the Eleco case raises particular concerns

Profiling refers to the process of inferring personal characteristics, preferences, and even psychological states through collected behavioral data. In the Eleco case, the problem isn't just that data was unencrypted — it's that the company actively used private conversations from children and family settings to build customer profiles with psychological judgments.

Such practices are strictly restricted in many countries. For example, the EU's General Data Protection Regulation (GDPR) classifies psychological inference as sensitive data processing, requiring a higher degree of consent and transparency; the U.S. Federal Trade Commission has also taken enforcement action against smart toy makers for children, citing violations of the Children's Online Privacy Protection Act (COPPA).

For hardware manufacturers, Eleco's lesson is: once a product targets children or families, data compliance standards rise significantly, and any "unprotected transcription" can escalate from a technical oversight into a legal and reputational crisis.

What the Three Threads Point To

The iPhone Duo buying rush, the Apple-OpenAI lawsuit, and the Eleco leak appear to be three independent news stories, but they actually point to the same thing: AI is simultaneously reshaping the boundaries of consumer electronics, corporate competition, and personal privacy — and the rules around those boundaries are far from established. AI agent buying tests market fairness, trade secret litigation tests talent mobility rules, and the AI pet vulnerability tests the most basic compliance floor.

This content is compiled from The Verge's Vergecast podcast (Thursday, October 15, 2026 episode).

One-line Summary

iPhone Duo pre-orders face AI agent buying, compounded by the Apple-OpenAI lawsuit and an AI pet data leak, reflecting the disorder at AI's boundaries.

Paragraph Summary

Drawing from The Verge's Vergecast podcast, the article connects three news stories: Apple's first foldable iPhone Duo opens "pre-pre-orders," but the real competitors are AI agents capable of ordering in milliseconds, leaving ordinary consumers virtually unable to compete fairly; the Apple-OpenAI trade secret lawsuit reaches a turning point as OpenAI promises not to use hardware secrets obtained from former Apple employees, timed just after a judge indicated a likely injunction; and AI pet toy Eleco is exposed for leaking millions of unprotected conversations, with the company using them to psychologically profile customers. All three threads point to one conclusion: AI is simultaneously reshaping consumer electronics buying rules, corporate intellectual property competition, and the privacy floor in the age of always-listening devices — while the corresponding rules remain far from established.