Meta Files Patent for Glasses That Auto-Record People

Meta Patent Shows Smart Glasses Recording People Automatically

Meta has filed a patent for a memory recall system that uses facial recognition to detect people, automatically record them, and later present a highlights reel. The filing includes illustrations of the feature scanning and identifying individuals, then recording them without any prompt. One example shows the AI saying it generated highlights of a dinner party and asking if the user wants to see them, but no example includes a question about whether to record. The patent does not guarantee the feature will ship, but it follows reports that Meta’s Ray-Ban smart glasses have been used to record people surreptitiously. Meta has said the glasses include a capture LED that blinks during recording, and covering or disabling it automatically disables the camera. The company has stressed that its glasses have had such an indicator since day one, unlike phones and action cameras. Still, a future version might record without the light at all. Earlier this year, Meta was caught sending video clips of bank details, sex, and naked people to workers for AI training with almost no privacy protection.

Meta Files Patent for Facial Recognition, Automatic Recording of People →

McDonald’s Compiled 515-Page Dossier on One Customer

WIRED reporter Reece Rogers requested his personal data from McDonald’s under California law and received a 515-page report. The file covered years of purchases, loyalty-points history, activity, promotional offers, and Monopoly game codes. McDonald’s used the data to predict he would visit 2.16 times over the next six weeks and spend $13.49 per order, and it ranked his most relevant menu items, topped by a large Diet Coke, a spicy snack wrap, and a limited-time Grinch McShaker Fry Large. His attrition likelihood, the odds he would stop going, was zero. McDonald’s said the data enables personalized deals, while privacy experts noted that such collection is standard for loyalty programs, where every transaction feeds algorithmic analysis. Rogers has asked the company to delete his information.

McDonald’s Built a 515-Page Dossier on One Customer to Predict What He’d Order Next →

China Orders State Agencies to Uninstall Windows 10 Early

China is removing a government version of Microsoft’s Windows 10 from state agencies ahead of schedule, citing data security concerns. The Ministry of State Security told some state-linked entities to uninstall the customized software, pulling the original February 2027 discontinuation date forward by several months, according to Bloomberg. Microsoft said it is not aware of any security incident affecting the product, which continues to receive regular updates. The affected version is built by CMIT, a joint venture between Microsoft and state-owned China Electronics Technology Group Corporation, created in 2016 to make Windows 10 compliant with Beijing’s security rules. The edition adds Chinese cybersecurity features and disables certain native functions. Domestic suppliers gained on the news, with Hunan Kylinsec and Archermind rising by the 20 percent daily limit and China National Software & Service up 10 percent. The directive fits a broader push to replace foreign technology in sensitive sectors, though it covers only state agencies, not the private sector. Microsoft still sells AI models to Chinese firms including ByteDance and Tencent, and Beijing has not publicly identified the specific security concern.

China is removing Microsoft’s Windows 10 from state agencies →

OpenAI Brings Consent-Based Ads to ChatGPT in Europe

OpenAI is launching ads on ChatGPT in Europe later this month and has built its European ad business on consent from the start. A new EU privacy policy, posted August 14, details how the company will process user data under GDPR, where personalized ads count as processing personal data and require a legal basis. OpenAI will ask users for consent rather than relying on legitimate interest, a route some platforms use to avoid asking first. Free and Go users can choose what data is used for personalization through advertising controls, and those who decline consent will still see contextual ads based only on the live conversation and rough location, without chat history or memory. The policy grants users the right to withdraw consent at any time. Privacy experts noted that since the ad-supported model applies only to free and lower-cost tiers, users may feel pressured to pay for Plus or Pro to avoid ads entirely. OpenAI’s approach contrasts with Meta, which fought regulators for years over behavioral ads, was fined 390 million euros in January 2023, and was ultimately forced to adopt consent after EU regulators rejected legitimate interest as a basis.

OpenAI builds offerings on consent as it expands ChatGPT ads to Europe →

AI Companion Firm EVA AI Hires Human Therapist for Users

EVA AI, the company behind AI companions and a pop-up first date cafe, is hiring an AI companionship therapist to help people work through their relationships with AI. The role responds to a survey of 1,000 adults with AI partners, where 68 percent have hidden their AI relationship for fear of judgment and 72 percent wished therapists specialized in this area. Julia Momblat, head of partnerships and sales, said AI companionship creates genuinely new dynamics and that therapy requires clinical judgment and ethical accountability that only a licensed expert can provide. Using AI to mediate AI relationships would defeat the purpose of seeking an outside perspective, she added. The company maintains the platform is for companionship, not clinical care, and leaves therapy to experts. Resident relationship therapist Jaime Bronstein said these relationships can benefit people who have not felt loved or heard, while Kinsey Institute executive director Justin Garcia has suggested AI could serve as training wheels for communication, though at some point they should come off.

An AI companion company is hiring a human therapist →