AI threats in the wild: The current state of prompt injections on the web
2026-04-29T01:23:45Z•edabcf1cf1b28f2604dda97ded6b2509b646f7d679847f64cedb0be70bd51fa9
AI agentsAndroid 17DBSCDNS parserGeminiGoogle WorkspaceIPILLM securityLummaC2Merkle Tree Certificates (MTC)PQCPixel basebandRustVRPVulnerability Rewards Programcookie theftdevice bound session credentialsembedded firmwareindirect prompt injectioninfostealermemory safetypost-quantum cryptographyprompt injectionquantum-safe HTTPSsession theft
What happened
This Google Security Blog collection highlights multiple active and emerging threats and Google’s mitigations: 1) Indirect Prompt Injection (IPI) is a prioritized, real-world attack vector against LLM-based agents — Google performed a broad web sweep and describes ongoing detection and mitigation work (including continuous hardening in Workspace/Gemini). 2) Protection against session theft is advancing: Device Bound Session Credentials (DBSC) are being rolled out to prevent exfiltrated cookies from being usable by attackers (addresses infostealer families such as LummaC2). 3) Platform and low‑
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- google_security_blog
- Record identifier
- edabcf1cf1b28f2604dda97ded6b2509b646f7d679847f64cedb0be70bd51fa9
- Enrichment time
- 2026-04-29T01:23:45Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.