Discover and rank the freshest highest-impact AI Security delta by freshness + technical impact + evidence strength + real-world exploitation + defensive value; exclude prior ARTEX/Cursor/LiteLLM identities before durable commit.
RUN AUDIT
سجل تدقيق مستقل يربط نتيجة البحث بالأدلة والتناقضات وحالة الالتزام والـReadback والمتابعة دون استنتاج حالة غير مسجلة.
Discover and rank the freshest highest-impact AI Security delta by freshness + technical impact + evidence strength + real-world exploitation + defensive value; exclude prior ARTEX/Cursor/LiteLLM identities before durable commit.
NEW_TOPIC
Research result: PASS
Selected Load/Mine 24/25. Verified malicious model-repository custom-code execution, miner/reverse-shell persistence, post-takedown recurrence, container-escape variant, official trust_remote_code semantics and strong exposure/attribution negative-space.
EVIDENCE
CONTRADICTIONS
لا توجد تناقضات مسجلة لهذه الجولة.
UNCERTAINTY
FOLLOW-UP
Authoritative counts of unique loaders, malicious-code executions, miner installs, persistence/backdoor success, container escapes and affected organizations are unknown.
Complete authoritative inventory of remaining copies, replacement accounts, mirrors and current C2/mining infrastructure is not established.
Operator identity and whether all campaign waves/linked repositories share one actor remain unproven.
No controlled ecosystem-wide baseline quantifies which executable model-repository behaviors are detected or missed by current hub/partner scanners.
AUDIT TIMELINE
لا توجد run_events تاريخية لهذه الجولة؛ تُعرض حالات Lifecycle أعلاه كما هي من سجل Run ولا يتم اختلاق أحداث بديلة.
PERSISTENCE / READBACK
COMMITTED · Readback PASS
SELF
Integrity: PASS · غير مسجل
Evidence 143 · Hypotheses 101 · Runs 30
Commit: PASS · Readback: PASS · Closure: PASS
PASS — AIS SSOT + atomic central commit/readback + identity/dedupe + Hot-Start v1.4 final readback. Load/Mine NEW_TOPIC; 9 canonical memory objects, 15 non-orphan edges, 4 open gaps, 1 watch; protected Core unchanged.
NEW — malicious Hugging Face/ModelScope model-repository campaign verified with custom Python execution, GPU mining, reverse-shell/persistence variants and post-takedown recurrence. Downloads do not equal infections; current repository state, realized victim denominator, universal scanner coverage and actor identity remain gated.