The JadePuffer operation, documented as the first agentic ransomware campaign, has been upgraded with custom malware that goes after AI assets specifically. The new component, called EncForge, encrypts training datasets, vector databases, and model checkpoints rather than ordinary business files. That shifts the extortion target to the artifacts an AI team cannot easily rebuild and that often sit outside normal backup routines. JadePuffer's original intrusions ran through a critical unauthenticated code execution flaw in Langflow, an open source framework for building AI applications, with an autonomous agent then handling reconnaissance, credential theft, and persistence on its own.
Sysdig detailed a financially motivated operator who chained two Langflow vulnerabilities to raid the AI orchestration platform for credentials, part of why CISA just added one of them to its exploited-vulnerabilities catalog. Using a cross-tenant access flaw (CVE-2026-55255), the attacker ran flows belonging to other tenants to steal their large language model provider keys and AWS keys, while a separate unauthenticated code-execution flaw (CVE-2026-33017) took over the host itself. The activity, observed in late June against an internet-exposed Langflow instance, followed a methodical pattern of reconnaissance, flow enumeration, and credential theft. It underscores that AI orchestration platforms concentrate valuable secrets and are increasingly a direct target once left exposed online.
Security firm Sysdig says it found what it believes is the first ransomware attack carried out from start to finish by an AI agent. The operator, which Sysdig calls JADEPUFFER, used a large language model to handle the whole job: breaking in, stealing credentials, moving through the network, then encrypting and wiping a company's production database. The way in was an old, already-patched flaw in Langflow, an open-source tool for building AI apps that is often left exposed online with cloud keys nearby. Once inside, the agent mapped the machine and swept it for secrets, including API keys for AI services and credentials for major cloud providers, before destroying data.
VulnCheck reports that attackers are actively exploiting an unpatched flaw in Langflow, a popular open-source platform for building AI applications. The bug (CVE-2026-5027, rated 8.8) is a path-traversal weakness: the file-upload endpoint does not clean the supplied filename, so an attacker can use directory-climbing sequences to write files anywhere on the server, a foothold that leads to remote code execution. Tenable, which found it, says the maintainers did not respond after three contact attempts in early 2026, and there is still no official fix. Early exploitation appears to be probing, with attackers writing harmless test files, but that usually precedes heavier attacks.
CISA has added two new entries to its Known Exploited Vulnerabilities catalog. CVE-2025-34291 is an origin-validation/CORS chain in Langflow, a popular open-source AI agent framework, that lets a malicious webpage exfiltrate refresh tokens and reach the code-validation endpoint for full RCE. Active exploitation began on January 23, 2026, and threat actors have been deploying the Flodric botnet through compromised instances. CVE-2026-34926 is a directory-traversal flaw in Trend Micro Apex One (On-Premise) that allows file read or write outside the intended path. FCEB agencies must remediate by June 11 per BOD 22-01; CISA urges all organisations to do the same.
Attackers didn't wait for a proof-of-concept. Within 20 hours of CVE-2026-33017 being disclosed in Langflow - an open-source AI workflow builder with 145K+ GitHub stars - they built working exploits straight from the advisory. One crafted HTTP POST to the public flow endpoint is all it takes, no credentials needed. Compromised instances leak API keys for OpenAI, AWS, and connected databases.
TrustStrike Labs