Signal is active with 4 mentions in latest observed window
Recommended action window: Monitor and triage in normal cycle
NVD description
PraisonAI before 4.6.78 fails to validate the caller-controlled dimension argument in the PGVector and Cassandra knowledge-store create_collection() backends. Although schema, keyspace, and collection-name identifiers are validated, the dimension value (declared as int but not enforced at runtime) is interpolated directly into the vector column of the generated CREATE TABLE DDL. A caller able to influence collection-creation dimensions can pass a string such as '3); DROP TABLE tenant_secrets; --' to inject SQL/CQL tokens into the statement executed by the database driver.
Priority
LOW
Exploitation
NONE
PoC
NONE
Patch
AVAILABLE
Momentum
NONE
🚨 CVE-2026-60090 — CVSS 9.8/10 ██████████ PraisonAI before 4.6.78 fails to validate the caller-controlled dimension argument in the PGVector and Cassandra... Severity: CRITICAL Patch now. #cybersecurity #CVE https://t.co/kjRnESRqYs
Post summary
A critical vulnerability (CVE‑2026‑60090) in PraisonAI’s PGVector and Cassandra modules has been disclosed, highlighting a missing validation for a dimension argument; a patch is now available.
🚨*CVE* CVE-2026-60090 PraisonAI before 4.6.78 fails to validate the caller-controlled dimension argument in the PGVector and Cassandra knowledge-store create_collection() backends. Althoug… https://www.cve.org/CVERecord?id=CVE-2026-60090 ----- Traducción: CVE-2026-60090 Pra… http://infoflow.cloud`
Post summary
The message announces CVE-2026-60090 impacting PraisonAI versions prior to 4.6.78, detailing a validation failure related to caller‑controlled dimension arguments in specific backends.
CVE-2026-60090 PraisonAI before 4.6.78 fails to validate the caller-controlled dimension argument in the PGVector and Cassandra knowledge-store create_collection() backends. Althoug… https://www.cve.org/CVERecord?id=CVE-2026-60090
Post summary
The text briefly announces CVE‑2026‑60090, outlining a validation vulnerability in PraisonAI’s knowledge‑store backends, and provides a link to the official CVE record, without mentioning exploitation, patches, or debunking.
🚨Critical - Two PraisonAI flaws: prompt-injection-to-root RCE and SQL injection (CVE-2026-61445 & CVE-2026-60090) PraisonAI, the open-source multi-agent framework, has two critical vulnerabilities fixed in 4.6.78. CVE-2026-61445 (CVSS 9.4): the AICoder component performs no path validation or command sanitization on LLM tool calls. A crafted prompt sent through the chat interface can write files to arbitrary filesystem paths and execute shell commands with root privileges - prompt injection escalating directly to full host compromise. CVE-2026-60090 (CVSS 9.8): the create_collection() function in the PGVector and Cassandra knowledge-store backends never enforces that the "dimension" argument is an integer. A string value is interpolated straight into the CREATE TABLE DDL, enabling SQL/CQL injection such as dropping tenant tables. Both require only low privileges over the network. If you run self-hosted PraisonAI with tool execution or a vector knowledge store, treat as urgent. 👉Upgrade praisonai to 4.6.78 or later.
Post summary
The notice announces two critical CVEs (CVE‑2026‑61445 and CVE‑2026‑60090) in PraisonAI, provides technical details and CVSS scores, and recommends upgrading to version 4.6.78 to mitigate the vulnerabilities.