Paper: LLM memory injection can persist across fresh sessions
Authors say all NIST attack goals were met on ChatGPT Instant 5.3, Claude Sonnet 4.6, and Gemini 3 Flash.
3 stories, newest first · one card per story
Authors say all NIST attack goals were met on ChatGPT Instant 5.3, Claude Sonnet 4.6, and Gemini 3 Flash.
A zero-parameter Memory Decision Layer cuts LLM hallucination from conflicting retrieved memories by about 56 percent.

Sophos found F5 BIG-IP APM malware that injects a PHP web shell into Apache's memory, evading disk scans, linked to exploited CVE-2025-53521.
Today 1.6K readers · 2.5K views · 780 reads · all time 189K views