#privacy-ai News & Analysis

9 articles tagged with #privacy-ai. AI-curated summaries with sentiment analysis and key takeaways from 50+ sources.

9 articles

AIBearisharXiv – CS AI · Jun 257/10

🧠

Erased, but Not Gone: Output Forgetting Is Not True Forgetting

Researchers demonstrate that machine unlearning methods that appear successful at the output layer—the standard evaluation metric—actually retain structured residual information in representation space compared to true retraining. This finding reveals a critical gap between apparent forgetting and genuine forgetting, suggesting current unlearning evaluations systematically overestimate effectiveness.

AIBullishCrypto Briefing · Jun 97/10

🧠

Apple unveils AFM 3 Core Advanced with 20 billion parameters for on-device AI at WWDC26

Apple announced the AFM 3 Core Advanced, a 20 billion parameter on-device AI model at WWDC26, marking a significant step in bringing advanced AI capabilities directly to consumer devices. The move underscores the industry's shift toward specialized hardware designed to support sophisticated AI processing without relying on cloud infrastructure.

AIBullishGoogle DeepMind Blog · Oct 237/104

🧠

VaultGemma: The world's most capable differentially private LLM

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.

AIBullisharXiv – CS AI · Jun 196/10

🧠

Techniques for Peak Memory Reduction for LoRA Fine-tuning of LLMs on Edge Devices

Researchers introduce memory optimization techniques for fine-tuning Large Language Models using LoRA on resource-constrained devices, achieving up to 28× peak memory reduction through quantization, efficient checkpointing, and token approximation methods. The work enables private model personalization on consumer hardware without compromising model quality.

🧠 Llama

AIBullisharXiv – CS AI · Jun 116/10

🧠

ASRU: Activation Steering Meets Reinforcement Unlearning for Multimodal Large Language Models

Researchers introduce ASRU, a machine unlearning framework for multimodal large language models that balances removing sensitive information with maintaining generation quality. The approach uses activation steering and reinforcement learning to achieve superior unlearning effectiveness while preserving model utility, demonstrating significant improvements on Qwen3-VL.

AI × CryptoBullishCrypto Briefing · Jun 86/10

🤖

Jon: Venice prioritizes user privacy over data exploitation, aims to be a household AI brand, and focuses on usability for non-crypto users | Bankless

Venice, an AI platform, is positioning itself as a privacy-focused alternative to centralized AI services, emphasizing user data protection and accessibility for non-technical audiences. The project aims to establish itself as a mainstream AI brand while maintaining crypto-native principles around privacy and decentralization.

AI × CryptoBullishBankless · Jun 46/10

🤖

Venice's Plan to Take on OpenAI

Venice is positioning itself as a competitor to OpenAI and Anthropic by offering private, uncensored AI inference aggregated across multiple models. The company plans to monetize this infrastructure by selling inference capacity to AI agents, creating a two-sided business model targeting both privacy-conscious consumers and emerging agent-based applications.

🏢 OpenAI🧠 Claude

AINeutralarXiv – CS AI · Jun 26/10

🧠

How Hard Can It Be? Hardness-Aware Multi-Objective Unlearning

Researchers introduce HAMU, a machine unlearning algorithm that removes the influence of specific training data while preserving model performance by quantifying the difficulty of balancing forget quality and retain utility through data similarity metrics. The approach offers theoretical guarantees and practical deployability for non-convex models, addressing a critical privacy and bias concern in machine learning.

AI × CryptoNeutralcrypto.news · May 126/10

🤖

Upbit listing puts privacy AI token VVV in Korea’s spotlight

Upbit, South Korea's largest cryptocurrency exchange, has listed Venice Token (VVV), a privacy-focused AI token, with multiple trading pairs including KRW, BTC, and USDT. The listing brings increased visibility to the privacy AI sector in Korea's competitive crypto market, though VVV's price declined 2.9% in 24 hours to approximately $17.15.

$BTC