Topic
data-privacy
Technology Amazon Can Use Twitch Content to Train AI Models Unless Users Opt Out
Twitch has added an opt-out toggle that lets streamers stop Amazon from using their content to train AI models. The change follows pushback from more than 16,000 creators, and WIRED reports the terms of service had not explicitly covered generative AI training. The case highlights the growing demand for training data and the consent gaps around user-generated content.
Technology McDonald's Built a 515-Page Dossier on a Customer. It Predicts He'll Never Leave
A WIRED reporter requested his personal data from McDonald's and received a 515-page file containing predictions about his future visits, spend, and loyalty. The report shows how a loyalty program's transaction data feeds predictive analytics, with an attrition likelihood of 0 and expected spend of $29.15 over six weeks.
Technology Apple challenges UK government again over encrypted data access order
Apple has confirmed it has launched a new legal complaint against the UK government at a court that handles objections to covert surveillance powers. The Financial Times reports the complaint is another challenge to a Home Office demand for backdoor access to encrypted Apple user data protected by Advanced Data Protection, according to BBC News. Privacy groups Privacy International and Liberty have welcomed the challenge.
Technology Study Finds Mobile Apps Marketed to US Troops Contain Chinese and Russian Code
A study by Purdue University, West Point, and Florida International University examined 220+ mobile apps marketed to US troops, finding that more than one in eight contained software from foreign adversaries like China and Russia. 64% had third-party SDKs, 40% collected more data than disclosed, and 12 apps included Huawei's HMS Core, posing risks for troop location tracking.
Technology How Consumer Fitness Trackers Are Quietly Predicting Pregnancy Earlier Than Tests
Fitness trackers like the Oura ring are alerting users to pregnancy through changes in heart rate, temperature, and HRV, days before standard tests. The phenomenon raises questions about data privacy and the predictive power of wearable biometrics.
Technology Meta's New AI Image Model Uses Public Instagram Photos by Default—Here's How to Opt Out
Meta launched its Muse Image AI model, allowing users to generate AI images using public Instagram profiles. Public accounts are automatically opted in; users must manually opt out to prevent their photos from being used. This raises privacy concerns for enterprise professionals managing brand presence.
Technology Meta Halts Worker Tracking for AI Training Amid Privacy Backlash
Meta has paused a company-wide initiative that tracked employee mouse clicks and keystrokes for AI training, following privacy fears and a petition signed by nearly 2,000 workers. The program, called the Model Capability Initiative, was halted after data was left potentially accessible to all employees.
From Privacy to Workflow Integrity: Communication-Graph Metadata Threat in Autonomous Agent Interoperability
A recent study published on arXiv formalizes the threat model for communication-graph metadata in autonomous agent interoperability protocols such as A2A and MCP. The research finds that while message content is protected, the graph of which agent contacts which, when, and how often can reveal pending workflows with high precision, enabling an adversary to act before the workflow completes. The paper argues this constitutes a workflow integrity risk rather than a mere privacy violation, and evaluates candidate transports to mitigate the leak.
Privacy-Preserving Text Sanitization for Distributed Agents via Disentangled Representations
Researchers propose DiSan, a privacy-preserving text sanitization framework that uses disentangled representations to separate task semantics from style identifiers. Experiments show it reduces personally identifiable information exposure by 20 times while maintaining 83% answer faithfulness on a multi-agent RAG benchmark, outperforming token-level masking.
MINT Demo 2 Framework Detects Training Data in Vision-Language Models With 90% Accuracy
Researchers introduced MINT Demo 2, a framework to determine if specific data was used to train vision-language models. The system achieves up to 90% accuracy and includes a web platform for auditing multiple model types, aiming to improve AI transparency and regulatory compliance.
New Auditing Framework Detects Synthetic Data Privacy Leaks Without Model Access
A new causal framework for auditing synthetic data detects privacy leaks by distinguishing true disclosures from phantom ones. It uses statistical hypothesis testing with holdout sets, requires no model access or canary insertion, and is orders of magnitude more efficient than shadow-model approaches.
Technology Smart TVs Become Nodes in AI Scraping Networks, Security Research Reveals
Security firm Include Security documents how Bright Data's residential proxy network uses consent-based SDKs embedded in consumer apps, including smart TVs, to route AI training data scraping traffic through home internet connections. The research highlights the unique advantages of connected TVs over mobile phones for this purpose, including always-on power, high-speed WiFi, and minimal user oversight.
Technology 1,000 Data Breaches Later, the Disclosure Lag is Worse Than Ever
Troy Hunt loaded the 1,000th breach into Have I Been Pwned, highlighting that disclosure lag times are worsening despite GDPR and CCPA. Examples include Carnival's 43-day delay and Zara's 45-day silence after ShinyHunters attacks, leaving victims uninformed for weeks.
Anti-Palantir Protests Erupt at UK NHS Conference Over Data Privacy Fears
About 80 protesters, organized by Pull the Plug, demonstrated outside the NHS ConfedExpo in Manchester against Palantir's NHS contract worth up to $440 million. The deal, which includes AI and data analytics, faces scrutiny over national security, data privacy, and political affiliations. The UK government is reviewing the contract ahead of a break clause next February.