MosaicLeaks
PulseAugur coverage of MosaicLeaks — every cluster mentioning MosaicLeaks across labs, papers, and developer communities, ranked by signal.
2 day(s) with sentiment data
-
Hugging Face explores AI agent security and CLI optimization · 3 sources tracked
Hugging Face is exploring new tools and concepts for AI agents, focusing on their security and operational efficiency. One area of research, highlighted by the "MosaicLeaks" project, investigates whether AI agents can m…
-
Hugging Face details vLLM updates, agent CLI, and data security research · 3 sources tracked
Hugging Face has released several blog posts detailing new developments and research. One post discusses the evolution of vLLM from version 0 to 1, focusing on pre-reinforcement learning accuracy. Another post introduce…
-
Hugging Face details vLLM, CLI for agents, and MosaicLeaks security
Hugging Face has released several blog posts detailing new developments in AI tooling and research. One post discusses the evolution of vLLM from version 0 to 1, focusing on its accuracy before reinforcement learning co…
-
New MosaicLeaks benchmark reveals AI research agents leak private data
Researchers have introduced MosaicLeaks, a new benchmark designed to evaluate the privacy risks associated with AI research agents that combine private local documents with external tools. These agents can inadvertently…
-
New benchmark reveals AI agents leak private data via query aggregation
Researchers have identified a significant privacy risk in AI agents that combine private documents with external tools, such as web searches. This risk, termed the "mosaic effect," occurs when individual queries seem in…