The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has directed federal agencies to patch an actively exploited vulnerability in the Langflow AI development framework, by July 10, 2026, after adding it to the Known Exploited Vulnerabilities (KEV) catalog. The flaw, tracked as CVE-2026-55255, is an authenticated insecure direct object reference (IDOR) flaw allowing attackers to access other users’ AI workflows by submitting malicious requests containing a victim’s flow ID, exposing sensitive data processed by those workflows while enabling unauthorized resource consumption. Sysdig researchers observed in-the-wild exploitation beginning on June 25, with financially motivated threat actors using the vulnerability to achieve remote code execution (RCE) and deploy second-stage malware, seeking to monetize compromised AI infrastructure through botnet activity and the theft of valuable cloud and large language model (LLM) credentials. The KEV addition continues a pattern of active targeting against Langflow, with CISA also highlighting exploitation of the previously disclosed CVE-2025-3248 missing authentication flaw (recently linked to the JadePuffer ransomware group’s theft of Langflow PostgreSQL databases) as well as the earlier CVE-2026-33017 code injection vulnerability. Separately, researchers have reported ongoing exploitation of CVE-2026-5027, a path traversal flaw that allows attackers to write arbitrary files to exposed Langflow servers, underscoring the growing focus on AI development platforms as attractive targets for financially motivated cybercriminals. CTIX analysts urge any affected organizations to ensure they patch the flaw immediately to prevent exploitation.