LLM-response data recoverable per query via LeftoverLocals (CVE-2023-4969) from un-scrubbed GPU local memory

Name: LLM-response data recoverable per query via LeftoverLocals (CVE-2023-4969) from un-scrubbed GPU local memory
License: https://aidatacenterguide.com/terms#intellectual-property

≈181 MBobserved

Value kind	observed — Reported measurements, counts, and specifications keep the precision and scope stated by their source; an exact specification is not treated as a range.
As of	2024
Source	Trail of Bits · Introductory findings: about 5.5 MB per GPU invocation and about 181 MB for each llama.cpp 7B-model query
Review	checking…review by 2026-10-07 · standard cadence
Recorded changes	last 2026-06-29
Claim id	llm-response-data-recoverable-per-query-via

Where the guide uses it

11.6 Multi-Tenant & Workload Isolation Security

← Full numbers register — every date-stamped figure in the guide, with revision history.