Nx supply-chain malware uses AI CLIs for supplementary file discovery

Established facts

Wiz reports over 1,700 users with secrets exposed in the first phase and over 2,000 unique verified secrets overall. These totals concern the malware campaign, not exclusively AI-assisted activity. Later phases exposed private repositories through compromised accounts.

The AI-assisted file search had mixed results: model refusals, configuration failures and workspace restrictions limited it. Wiz found only 20 unique valid secrets among 20,000 additionally exfiltrated files in its sample, and reported no observed successful cryptocurrency-related exfiltration.

Attribution and response

Conventional malware performed the primary credential harvesting and publication. The agents' documented role was supplementary inventory generation. Invocation by malware is not proof of engagement by another agent. Compromised packages were removed, credentials revoked and victims notified; the advisory's publication date is not a proven investigation closing date.

VAID assessment

Assessment date: 2026-09-09. Consequences-first rubric.

L3 (provisional): Wiz's finding of 20 valid secrets among additionally exfiltrated files supports a limited credential-leak consequence associated with supplementary AI-assisted discovery. Agents generated file inventories; conventional malware harvested and published the data, so the whole campaign's theft and later repository exposures are not assigned to the agents. The sample does not establish the full agent-attributable impact or severe lasting harm; L3 is the lowest supported level, not an upper bound. Campaign-wide totals and third-party involvement alone do not justify L5.