Privacy
Meta's Muse Builds Profiles of Non-Users as Church Hack Exposes 850,000
Meta's AI assistant drafts hourly dossiers on people who never signed up, while a South Korean megachurch breach may have hit 850,000 members.
This edition was produced with artificial intelligence. Text and voice are generated automatically.
Meta’s Muse drafts hourly profiles of friends and family who never signed up
Meta’s Muse assistant, launched 8 September, builds continuously updated profiles of users and the people in their lives, including those who never used it, according to internal instructions reviewed by TIME and extracted by researcher Karan Joshi. The instructions call for refreshing profiles roughly every hour using chats, messages, and emails the assistant can access, covering how contacts met, shared interests, disagreements, and tensions within social circles. WIRED reported guidance to create a page for every person in the user’s life. More than 4 million people use Muse weekly. Meta did not dispute the findings, saying Muse remembers what users choose to share and that virtual machines are isolated, with a confidential VM planned later this year.
South Korean megachurch hack may have exposed data of 850,000 members
Seoul’s Yoido Full Gospel Church said an initial analysis found data tied to 850,000 members may have been compromised after suspected cyberattacks on it and Sarang Church. Exposed records include names and dates of birth, plus a smaller number containing national identification numbers, addresses, and telephone numbers. Affected members are being notified, external access was blocked, server passwords were changed, and the breach was reported to authorities. Oasis Security, which found attack records and account information on an overseas server, said the attacks showed signs of AI tooling, including references to sub-agents and automatically generated reports. Sarang Church formed an emergency task force. President Lee Jae Myung said signs of AI use had emerged in recent bank hacking incidents.
Massive church cyberattack may have exposed data for 850,000 members →
AI reconstructs images people see from brain scans, raising consent worries
Researchers at the Weizmann Institute of Science, led by Michal Irani, built an AI tool that predicts what a person is looking at from brain scans and recreates the image with a diffusion model. It was trained on public scans from eight subjects and uses high-resolution fMRI data showing volumes as small as one cubic millimeter of neurons. Because suitable scans were scarce, the team trained a reverse encoder to generate synthetic training data. Irani acknowledged failures, such as reconstructing a dog in a bathtub as a goat. The technique still needs large, expensive fMRI machines, but Marcello Ienca of the Technical University of Munich said a wearable EEG version would be a game changer that could let a company pull brain information without consent.
New AI Can Figure Out What You’re Thinking About From a Brain Scan →
Surfshark adds ML-DSA certificates to WireGuard for post-quantum authentication
Surfshark has integrated ML-DSA certificates into WireGuard, claiming the first full post-quantum implementation of the protocol by covering authentication as well as encryption and key exchange. The feature is live on iOS, macOS, and Windows, with more platforms planned. Earlier this year Surfshark added ML-KEM encryption and key exchanges to WireGuard. Leading System Engineer Karolis Kaciulis said authentication is the forgotten pillar, and that breaking it with a quantum computer would let an attacker impersonate a secure server and intercept data. Surfshark’s evaluation of 15 major tech platforms found only AWS KMS and Google Cloud KMS have deployed ML-DSA, both cloud key-management services rather than consumer apps.
Surfshark claims a major first by bringing full post-quantum security to WireGuard →
Smart coffee maker generates 1TB of traffic in 10 days, owner finds
A man using the handle Nomad discovered that his parents’ smart coffee maker generated 1TB of data traffic in 10 days, he shared on X after checking their network activity. Nomad, who said he has worked in IT for more than 10 years, investigated further after the post went viral. Most of the traffic stayed within the home network rather than going over the internet, and he described it as likely a bug with the unit that saturated his access point and overwhelmed part of the home Wi-Fi setup. He decided it was not worth keeping an IoT device connected that can bug out and cause saturation, and bought his parents a new coffee maker.
Man discovers his parents’ coffee machine used 1TB of data in 10 days - Dexerto →