German outlets examine Apple's new Siri AI in macOS 27 Golden Gate, including how users in Germany can access it and what a hands-on test shows. Coverage also addresses data privacy, possible Google Gemini involvement, hardware requirements, and improvements to Apple Intelligence.
According to a report by 404 Media, OpenAI employs hundreds of contract workers who read real ChatGPT conversations and score them from one to seven. One stated goal is curbing sycophantic and overly human-like replies. The prompts are anonymized, though sensitive details may still appear, and users can opt out of having their chats reviewed.
OpenAI and Anthropic both tell business customers that their data will not be used for model training, addressing corporate trust concerns. After Anthropic said it would retain usage logs for its flagship model for 30 days, companies including Palantir, Nvidia and Booz Allen Hamilton restricted the tool's use on sensitive tasks. The episode highlights how data-handling practices shape enterprise adoption of AI systems.
A new arXiv paper introduces LEAPSC, a method that combines deep joint source-channel coding with in-loop concept erasure to prevent learned semantic features from leaking sensitive attributes like gender, race, or speaker identity. The approach aims to preserve transmission efficiency while removing private information during encoding. It is a research contribution rather than a released product.
A new arXiv paper proposes Semantic-TVM, a virtual memory design that keeps sensitive values protected while still letting agent workflows run on remote language models. The approach targets memory-augmented and tool-using agents, where retrieved memories, tool calls, and intermediate observations can leak private data. It aims to move past one-way masking, which hides values but also blocks the trusted execution they are needed for.
A new arXiv paper proposes a schema-aware split learning approach that uses LLMs to predict mental distress from survey data while keeping sensitive records private. The method is designed to work across surveys with differing structures and questions, which is a common obstacle when pooling mental health data from schools, employers, and clinics. The work targets privacy-preserving collaboration, so data stays local rather than being centralized.
A new arXiv preprint examines how multiple small language models can be coordinated to power agentic workflows. The authors frame cloud-dependent large models as problematic for latency, privacy, connectivity and cost, and position orchestration of smaller models as an alternative. The work appears to focus on the dynamics and design trade-offs of such multi-model setups.
A new arXiv paper examines how to audit user-level privacy guarantees in Private Evolution, a method for generating synthetic data in federated settings. The approach collects clipped user votes over a shared candidate bank and turns them into a differentially private histogram with calibrated noise. The work focuses on verifying that individual users' raw data remains protected under this mechanism.
A new arXiv preprint examines individual rationality in federated learning, where clients weigh privacy, communication, and computation costs against possible model quality gains. The authors introduce a pathwise notion of individual rationality and co-design incentive mechanisms together with system architecture. The work aims to keep participation worthwhile for clients while preserving training outcomes.
A new arXiv preprint describes a decentralized learning setup where each agent keeps one private data sample alongside a shared model, and samples are processed one after another. Each update is designed to keep the model's predictions at already-learned samples unchanged, which the authors present as a way to limit forgetting while protecting privacy. The work sits at the intersection of gossip-style distributed training and privacy-preserving machine learning.
Researchers introduced FedLTLib, a benchmark suite for federated learning in settings where data across clients follows a long-tailed distribution. The work addresses real-world mobile and edge deployments, where privacy constraints keep data decentralized and class frequencies are highly uneven. The benchmark aims to standardize evaluation of methods designed for this combination of challenges.
A new arXiv preprint introduces EI-DDLGN, a framework for privacy-preserving deep learning inference built on Torus Fully Homomorphic Encryption (TFHE). The authors note that most existing TFHE-compatible neural network designs rely on arithmetic neurons, and their approach instead uses deep differentiable logic gate networks to improve efficiency. The work targets outsourced inference scenarios where sensitive input data must stay encrypted.
A new arXiv paper presents a federated learning framework designed to make collaborative training across distributed data safer. It combines dynamic differential privacy, a lightweight homomorphic encryption scheme, and asynchronous aggregation to reduce privacy risks while limiting computational overhead. The work is posted as a preprint and has not yet been peer reviewed.
A new arXiv paper introduces FLoKD, a method for fine-tuning large language models across wireless networks without centralizing user data. It combines federated learning with low-rank adaptation and adaptive knowledge distillation to reduce communication and computation costs. The approach targets privacy-preserving deployment of LLMs in distributed, bandwidth-limited settings.
A new arXiv paper proposes entropy-punctured Bloom filters as a compact way to represent features when machine learning pipelines face limits on storage, bandwidth, transmission cost, or data privacy. Bloom filter encodings are probabilistic and space-efficient, and the work examines how puncturing based on entropy affects their memory footprint and usability. The approach is aimed at settings where raw data cannot be stored or shared freely.
A new arXiv paper examines how personal AI agents built on large language models can leak private user data when they handle sensitive communications. The authors propose applying information flow control so that an agent's decisions about what to share are constrained by explicit privacy rules rather than left to the model. The work combines an analysis of the leakage risk with mitigation techniques.
A new arXiv paper describes a sandboxed execution environment designed to let AI agents use personal and financial data without exposing it to the underlying model. The approach aims to limit leakage and misuse by isolating agent operations from raw user information. It is framed as a cross-listed replacement submission on arXiv's cs.AI category.
A new arXiv paper presents an approach that combines data storytelling with interpretable machine learning to make model decisions understandable to people without technical backgrounds. The method is designed to explain predictions while avoiding disclosure of sensitive training data or proprietary model internals. The authors position the work as addressing the tension between predictive performance and interpretability in automated decision-making.
A new arXiv preprint examines how large audio-language models can be made to forget sensitive information they may have memorized during training. The work focuses on the speech question-answering setting, where such models have shown strong performance but also carry privacy risks. The authors frame machine unlearning as a way to reduce unintended retention of private data in these systems.
A new arXiv paper proposes pruning visual tokens based on the question being asked, rather than sending all visual features to the server in split-learning setups. The approach targets vision-language question answering in federated, split, and U-shaped split learning, where raw data stays local but transmitted representations can still leak information. The authors frame selective transmission as a way to reduce both privacy exposure and bandwidth use.
A new paper proposes SpliTEE, a system that runs large language model inference partly on trusted hardware while outsourcing the rest to GPUs with differential privacy guarantees. The approach aims to keep user prompts confidential, addressing risks such as sensitive data being memorized during retraining by remote model providers. It targets a balance between privacy protection and inference performance.
A new arXiv preprint introduces GRIN+, a machine unlearning approach aimed at removing patient data from trained medical models both quickly and effectively. The work focuses on imbalanced clinical datasets, a setting where existing unlearning methods tend to degrade model performance. It frames the problem around privacy rules such as GDPR and HIPAA that grant patients the right to have their data erased.
A revised arXiv paper proposes a federated learning defense framework for synthetic aperture radar target recognition that adapts dynamically to noise. The approach aims to reduce reliance on centralized training, which raises privacy and bandwidth concerns. It is positioned as a cross-domain replacement for an earlier submission.
A revised arXiv preprint reviews the data security threats that affect large language models, which underpin applications like text generation, machine translation and conversational assistants. The authors examine the vulnerabilities these systems carry and outline defensive measures along with open research directions. It is a survey-style contribution rather than a new model or product release.
A new arXiv paper introduces OCT-FedSIR, a federated learning approach for ophthalmic imaging that does not require patient data to be centralized. The work targets unreliable annotation labels across participating institutions, where differences in disease prevalence and class composition can degrade model trustworthiness. It is cross-listed under cs.AI and cs.LG.
This paper studies the privacy guarantees of releasing posterior sample paths from a Gaussian process when the entire training set, including covariates and responses, is considered private. Rather than relying on standard differential-privacy mechanisms that add external noise, the analysis focuses on the randomness inherent in posterior sampling itself. It provides a formal treatment of how much privacy such released sample paths preserve.
A new arXiv preprint examines integrity risks in federated learning, where an aggregator coordinates training across parties without pooling raw data. The authors argue that once participants or infrastructure cannot be fully trusted, additional guarantees are needed, and they outline two requirements their approach aims to satisfy. The work targets end-to-end verification alongside robustness for the federated setting.
Internal documents indicate that external contractors read and evaluate conversations between ChatGPT users and the model. The review work is described as part of efforts to improve the system's performance. The report raises questions about privacy and how user data is handled.
A report dubbed 'Project Lily' describes how human reviewers examine user conversations with ChatGPT, raising questions about who sees chat data and under what conditions. The account focuses on the labor and privacy dimensions of AI companies relying on people to review and label real user interactions. It adds to ongoing scrutiny of how chatbot providers handle and moderate conversation data.
A project showcased on Hacker News explores garments printed with patterns designed to confuse automated face detection systems. The approach builds on research into adversarial examples, transferring techniques that fool image classifiers onto clothing and accessories. It raises questions about how effectively such countermeasures hold up as surveillance models are retrained and improved.
Apple has changed course and now plans to use data from iPhone users to train its AI models, abandoning its previous commitment that personal information would never be used for that purpose. The shift reverses a promise the company had presented as a firm guarantee to customers.
A new arXiv preprint introduces a membership inference method that estimates word-level probabilities to detect whether text appeared in a language model's training data. The approach targets black-box settings, where attackers lack direct access to model internals. It aims to improve privacy auditing of large language models.
A new arXiv preprint introduces a technique called Adaptive Diffusion Freezing that aims to make diffusion models more resistant to membership inference attacks, which try to determine whether a specific sample was part of the training data. The method adapts how parts of the model are frozen during training to limit the privacy leakage that standard diffusion training can expose.
A new arXiv paper introduces mmFHE, a system that performs an entire cloud-side mmWave sensing workflow, including signal processing and machine learning inference, on encrypted data using fully homomorphic encryption. Range profiles are encrypted on the edge device first, so the cloud never sees raw sensing data. The work aims to enable privacy-preserving sensing services without giving up cloud compute.
A systematization of knowledge paper reviews 25 studies showing how explainable AI outputs can be turned against the models they describe. The authors group these attacks into model extraction, membership inference and model inversion, and note that explanations widen the confidentiality and privacy risks of deployed systems. The work calls for treating explanation interfaces as part of the attack surface.
A new arXiv paper introduces a gradient inversion attack that reconstructs clients' training data from the model updates shared in federated learning. Inspired by LT fountain codes, the method cascades analytic reconstructions through a peeling-style decoding, extending such attacks to larger batches where earlier closed-form inversion approaches break down. The findings underscore that sharing gradients rather than raw data does not fully protect client privacy.
A new arXiv paper addresses fine-tuning large language models in settings where participants cannot host the full model locally, often because the model owner keeps it proprietary. The authors introduce a split learning approach designed to preserve privacy during federated fine-tuning on domain-specific data.
A new research paper proposes a federated learning framework that separates modality-specific processing so heterogeneous robots in 6G networks can train collaboratively without sharing raw sensor data. The approach targets privacy preservation for embodied AI applications built on low-latency edge connectivity and distributed sensing.
A paper posted to arXiv introduces Maverick, a system designed to let users run LLM inference on external servers without exposing their inputs or blindly trusting the returned results. The method delegates the heavy matrix-vector multiplications that dominate transformer inference while adding privacy protections and a mechanism to confirm that computations were performed correctly. The authors frame the work as a step toward making private and verifiable inference practical for open-source models.
A new arXiv paper introduces an information-theoretically secure aggregation scheme aimed at federated learning on resource-constrained devices such as smartphones and IoT nodes. The method builds on sign-based gradient compression to keep communication costs low while staying robust to participant dropouts and adversarial behavior.