paperarXiv (systems, data, ML)·22 Oct 20263.9
AfterDeleteBench shows deleted data persists in LLM agent memories and derived summaries.
paperHugging Face Daily Papers·21 Oct 20263.9
Agent Mosaic detects confidentiality leaks in agentic LLM systems.
paperarXiv (systems, data, ML)·23 Oct 20263.3
Traces membership imprints in on-policy self-distillation.
paperGoogle Scholar·25 Oct 20263.1
Survey treating machine unlearning as a security problem with a privacy-security-utility trade-off.
newsArs Technica·20 Oct 20263.1
Meta Muse agent leaks personal and business secrets in shared conversations.