The Cybersecurity and Infrastructure Security Agency added CVE-2026-55255, an authorization bypass in the open source AI orchestration tool Langflow, to its Known Exploited Vulnerabilities catalog on July 7, giving federal agencies until July 10 to patch. The flaw is an insecure direct object reference in Langflow’s /api/v1/responses endpoint: an authenticated attacker can supply another user’s flow ID and execute that user’s flow as if it were their own, no further privileges required. It affects Langflow versions before 1.9.2 and carries a CVSS base score of 8.4, according to the vulnerability’s National Vulnerability Database entry.
Langflow is not a household name the way Adobe or Citrix are, and that is exactly why this addition matters. It shows up in the same KEV batch as a decades-old Adobe ColdFusion path traversal flaw that CyberTech covered as part of a wider collapse in patch windows for edge and application platforms. CISA’s catalog does not distinguish between legacy enterprise middleware and a two-year-old AI workflow builder that data science and engineering teams have adopted to chain large language model calls together. Both got the same treatment: added to KEV, both given the same three-day remediation deadline.
The original insight for security leaders is less about Langflow specifically and more about what its presence in KEV signals for asset inventories. Security teams that maintain vulnerability management programs scoped to traditional enterprise software, and treat developer-adopted AI tooling as shadow IT outside that scope, are going to keep missing catalog additions like this one until they extend discovery to cover self-hosted AI orchestration platforms the same way they already cover web servers and VPN appliances. An IDOR that lets one authenticated user run another user’s automated flow is a meaningful risk in any environment where Langflow instances are exposed to more than a single trusted team, since flows in these tools frequently hold embedded API keys and connections to production data sources.
Source: NVD
