42 articles tagged with #data-protection. AI-curated summaries with sentiment analysis and key takeaways from 50+ sources.
Researchers introduce RULER, a verification framework that detects machine unlearning failures at the representation level rather than just output metrics. The study reveals that popular unlearning methods pass traditional evaluation tests yet still retain encoded information about forgotten data in their internal representations, highlighting a critical gap in current verification protocols.
A comprehensive survey examines Pretraining Data Exposure (PDE) in large language models, unifying two previously isolated research areas—membership inference and data contamination—to assess whether specific data appeared in LLM training datasets. The work formalizes exposure levels, reviews attack and defense mechanisms, and highlights privacy and evaluation integrity risks as model sizes and training data scales continue to grow.
Venice AI applies cypherpunk principles to artificial intelligence inference, building privacy protections into AI systems rather than treating it as an afterthought. The project draws philosophical parallels to the cypherpunk movement's core belief that privacy must be architecturally embedded, not granted by benevolent actors.
Google has detected hackers actively using AI to enhance cyberattacks and breach computer systems, confirming long-standing security concerns about weaponized artificial intelligence. The discovery signals that the predicted convergence of AI capabilities with malicious intent has moved from theoretical risk to operational reality, potentially expanding the threat landscape for individuals and organizations worldwide.
Google Chrome has quietly installed a 4GB on-device AI model while simultaneously removing privacy disclosures that previously promised to keep user data off Google's servers. This move raises significant concerns about transparency and the erosion of privacy protections in mainstream browsers.
Researchers propose RPSG, a novel method for generating synthetic data from private text using large language models while maintaining differential privacy protections. The approach uses private seeds and formal privacy mechanisms during candidate selection, achieving high fidelity synthetic data with stronger privacy guarantees than existing methods.
Researchers have developed presidio-hardened-x402, an open-source middleware that filters personally identifiable information from AI agent payment requests using the x402 protocol before data reaches payment servers or centralized APIs. The tool achieves 97.2% precision in detecting PII with minimal latency, addressing a critical privacy gap where payment metadata is currently transmitted without data processing agreements.
Researchers present Opal, a private memory system for personal AI that uses trusted hardware enclaves and oblivious RAM to protect user data privacy while maintaining query accuracy. The system achieves 13 percentage point improvement in retrieval accuracy over semantic search and 29x higher throughput with 15x lower costs than secure baselines.
Meta faces accusations of concealing privacy facts about Ray-Ban smart glasses after workers reported viewing footage of people in bathrooms. The allegations raise serious concerns about user privacy and data handling practices for wearable AI devices.
Meta faces a lawsuit over privacy concerns regarding its AI smart glasses, with allegations that the company's marketing promised user control while subcontractors were actually reviewing customer footage including sensitive content. The legal action centers on discrepancies between Meta's privacy promises and actual data handling practices.
Researchers introduce WARP, a new defense mechanism for machine unlearning protocols that protects against privacy attacks where adversaries can exploit differences between pre- and post-unlearning AI models. The technique reduces attack success rates by up to 92% while maintaining model accuracy on retained data.
Researchers propose Partial Model Collapse (PMC), a novel machine unlearning method for large language models that removes private information without directly training on sensitive data. The approach leverages model collapse - where models degrade when trained on their own outputs - as a feature to deliberately forget targeted information while preserving general utility.
Researchers introduce Multi-PA, a comprehensive benchmark for evaluating privacy risks in Large Vision-Language Models (LVLMs), covering 26 personal privacy categories, 15 trade secrets, and 18 state secrets across 31,962 samples. Testing 21 open-source and 2 closed-source LVLMs revealed significant privacy vulnerabilities, with models generally posing high risks of facilitating privacy breaches across different privacy categories.
Researchers have developed AudAgent, an automated tool that monitors AI agents in real-time to ensure they comply with their stated privacy policies. The tool revealed that many AI agents powered by major providers like Claude, Gemini, and DeepSeek fail to protect highly sensitive data like SSNs and violate their own privacy policies.
Researchers demonstrate that large language models can successfully deanonymize pseudonymous users across online platforms at scale, achieving up to 68% recall at 90% precision. The study shows LLMs can match users between platforms like Hacker News and LinkedIn, or across Reddit communities, using only unstructured text data.
OpenAI is resisting the New York Times' request for access to 20 million private ChatGPT conversations, while simultaneously implementing enhanced security and privacy protections for user data. This legal dispute highlights growing tensions over data privacy and corporate access to AI conversation logs.
VaultGemma represents a breakthrough as the most capable large language model trained from scratch using differential privacy techniques. This development advances privacy-preserving AI by demonstrating that sophisticated models can be built while maintaining strong data protection guarantees.
OpenAI is challenging a court order from The New York Times that would require indefinite retention of ChatGPT and API user data. The company is fighting the demands to protect user privacy while addressing legal requirements and maintaining data protection commitments.
The article discusses the development of encrypted large language models using Fully Homomorphic Encryption (FHE) technology. This approach would allow AI models to process data while keeping it encrypted, potentially addressing privacy concerns in AI applications.
Researchers introduce Shadow Unlearning, a privacy-preserving machine unlearning method that removes training data influence from LLMs without exposing sensitive information to attacks. The Neuro-Semantic Projector Unlearning (NSPU) framework achieves this while maintaining model performance and is 10x more computationally efficient than existing approaches.
diffGHOST is a new conditional diffusion model that synthesizes mobility trajectories while preserving privacy through latent space segmentation. The approach addresses a critical gap in existing generative models that lack formal privacy guarantees despite handling sensitive personal movement data.
Researchers propose INO-SGD, a novel algorithm addressing the utility imbalance problem in individualized differential privacy (IDP) machine learning systems. The algorithm strategically down-weights sensitive data batches to prevent underrepresentation of privacy-protected subsets, improving model performance for high-privacy users while maintaining differential privacy guarantees.
Researchers propose PPRoute, a privacy-preserving framework for LLM routing that uses Secure Multi-Party Computation (MPC) to protect user data while dynamically selecting between model providers. The system achieves 20x speedup over naive MPC implementations through optimized encoder inference, multi-step model training, and an efficient Top-k algorithm, maintaining routing quality without sacrificing privacy.
This academic paper examines how AI and data science practices can paradoxically increase vulnerability of subjects they aim to protect, using a case study of computer vision analysis of children in monetized YouTube content. The authors develop an ethics protocol identifying four critical decision points—dataset design, operationalization, inference, and dissemination—where technical choices create vulnerabilizing factors including exposure, monetization, narrative fixing, and algorithmic optimization.
Commvault has launched AI Protect, a governance solution that provides rollback capabilities for autonomous AI agents operating in cloud environments. The platform addresses critical risks posed by AI systems that can independently delete files, access databases, modify infrastructure, and alter security policies without adequate oversight or recovery mechanisms.
