MemPrivacy introduces a reversible, fine‑grained privacy layer for edge‑cloud agents that keeps sensitive data off the cloud while preserving the agent's long‑term personalization, outperforming OpenAI's privacy‑filter by over 50 % F1 score and incurring less than 1 % utility loss.
The rapid democratization of DeepFake technology lowers creation barriers, turning ordinary face photos into tools for malicious content, privacy breaches, and potential black‑market abuse, prompting urgent personal safeguards and regulatory action.
Two open‑source projects—BabelDOC, a Python‑based PDF translator that retains original formatting using AI models, and OneAIFW, a Zig‑and‑Rust local AI firewall that anonymizes sensitive data before LLM queries—offer practical, privacy‑preserving solutions for researchers and developers.
GitHub Copilot once auto‑completed a seemingly real Chinese ID number for a public figure, sparking concerns that large language models can unintentionally expose personal data learned from public web sources, highlighting privacy and security challenges in AI‑driven code assistants.
GitHub Copilot unexpectedly generated a fabricated ID for Bilibili CEO Chen Rui, sparking concerns about how large language models trained on public data can inadvertently expose synthetic personal information and highlighting broader privacy and ethical issues in AI‑driven code assistants.
