Tag
A wave of privacy-focused gay dating apps like MeetMarket are emerging as alternatives to Grindr and Sniffies, emphasizing decentralized identity, end-to-end encryption, and ad-free experiences to address user dissatisfaction with data monetization and bots.
This paper proposes ManiF-SMC, a method for approximate machine unlearning that operates entirely in the representation space by pushing erased samples away from their original learned manifold representation toward their nearest semantic neighbors in the retained data, using a margin-based triplet loss guided by a self-mode-connectivity module for adaptive margins.
US tech firms Microsoft and Meta have shared the names of Dutch civil servants and academics working on European tech regulation with a US Senate committee, potentially exposing them to travel bans or sanctions. The Dutch cabinet has described the situation as extremely worrying and has raised the issue with the US ambassador.
Bert Hubert explains why governments are heavily reliant on Palantir software, citing their exceptional hands-on integration support and accounting tricks that bundle consultant services with licenses, and calls for developing replacement software with European values.
The article highlights the disconnect between the widespread hype about AI adoption on social media and the actual challenges faced in corporate environments, such as poor data infrastructure, privacy restrictions, and unrealistic management expectations.
A study found that all nine workplace monitoring apps analyzed share employee data with third parties like Facebook and Google, raising serious privacy concerns.
A developer built a fully local AI workflow using 3 Mac Minis and 9 AI agents, achieving zero cloud service bills while earning $10k/month.
A report from the Electronic Privacy Information Center (EPIC) reveals that major data brokers and AI companies, including Google, Meta, and OpenAI, use deceptive design patterns in opt-out forms that prevent consumers from effectively opting out of data sale and sharing.
Research reveals a market where cheap Claude API access is achieved via stolen identities, deepfaked KYC, and smaller models masquerading as Claude, with all data being logged permanently—posing significant security and privacy risks.
The article argues that user data is a massive untapped market (~$5T over 10 years) but remains uncaptured due to privacy regulations and user resistance; the only effective model is trading data for personalization rather than cash.
The article explores the tradeoffs companies face when deciding between building custom AI models and using APIs from providers like OpenAI or Anthropic, focusing on cost, data privacy, performance, and long-term control.
A Dutch suicide prevention hotline was found to share sensitive visitor metadata with Google and Microsoft without proper consent, leading to the suspension of tracking tools and potential GDPR violations.
The article discusses the decision to migrate from GitHub to self-hosted Forgejo, citing concerns over data ownership, reliability, and AI data collection practices. It highlights similar moves by the Dutch government and details the technical setup of a personal Forgejo instance.
The Internet Cleanup Foundation launched SecurityBaseline.eu, a platform auditing the cybersecurity posture of European governments, revealing widespread issues such as illegal tracking cookies and poorly encrypted emails.
The article argues that while local AI models are accessible, true agent ownership requires local, inspectable memory systems rather than vendor-controlled cloud storage. The author advocates for tools like MemOS Local and Hermes Agent to maintain execution traces and learned skills locally for better control and debuggability.
A security vulnerability has been identified in recent Android versions that allows any installed app to potentially leak certain network traffic.
Palantir is reportedly set to receive unrestricted access to UK National Health Service patient data, raising significant questions about data privacy and the role of private tech firms in public healthcare infrastructure.
Princeton's Center for Information Technology Policy analyzes the 'Make America AI-Ready' SMS course by the US Department of Labor, highlighting its accessibility while critiquing significant contradictions in its data privacy advice.
This paper systematically investigates unlearnable examples under diverse training paradigms, revealing that pretrained weights weaken existing methods, and proposes Shallow Semantic Camouflage (SSC) to maintain unlearnability by generating perturbations in a semantically valid subspace.
Users are raising concerns about Weights & Biases' new Master Service Agreement, which grants the company broader rights to use customer data, including ML models, for product improvement and AI feature development.