Actively exploited AI attacks
Attacks on or using AI systems that are being exploited in the wild. Every entry names its sources. Confirmed: CISA KEV or the vendor documents exploitation. Observed (research): seen in the wild by a named research team, not yet in KEV or acknowledged by the vendor. Unconfirmed: community or press reports only (e.g. Reddit, X). See Methodology.
Observed (research)
Blind prompt injection against exposed AI agent frameworks, observed in the wild (Wiz honeypots)
Source attribution: Wiz Research
Confirmed
Critical (CISA KEV)
Critical (CISA KEV)
LiteLLM MCP auth bypass (CVE-2026-59822) exploited in the wild; on CISA KEV since Sept 2, 2026
Source attribution: CISA KEV, GitHub Advisory, Wiz Research
Confirmed
Critical (CISA KEV)
Critical (CISA KEV)
LiteLLM MCP test-endpoint command injection (CVE-2026-42271) exploited to drop cryptominers
Source attribution: CISA KEV, GitHub Advisory, Wiz Research