AI Governance Institute
← News
Enforcement2026-11-17

FBI and AWS Warn ARTEX Is Automating Credential Theft and KYC Fraud at Scale

What happened

The FBI and AWS Chief Information Security Officer CJ Moses published separate alerts on November 14, 2026. The alerts detailed active use of ARTEX, an open-source AI-powered penetration testing framework, to automate attacks against critical sectors. Amazon's MadPot honeypot sensors detected two ARTEX campaigns. The first used AI agents to find and steal access keys for AI platforms. Those keys were fed back into ARTEX to power further searches. Targeted platforms included LLM gateway and proxy services: LiteLLM, One API, and New API. AI application builders Dify, Langflow, and Ollama were also targeted. The second campaign targeted two cryptocurrency exchanges. It used real-time AI-generated deepfake videos and forged ID documents to defeat identity verification checks. The vendor's server-side verification rejected the deepfake attempts in every trial. However, the agents came close to meeting frame-rate and positioning requirements. The FBI noted that ARTEX was used in the ARTEX multi-agent campaign that hit 200 organizations in a state-linked attack. It also cited the September campaign that compromised more than 2,250 Microsoft 365 tenants via a ConnectWise Cloud Backup supply chain attack. The original ARTEX developer removed the GitHub repository in October. Despite this, Censys identified 1,465 hosts running an exposed ARTEX console as of November 4. The FBI noted total deployments are likely significantly higher. Both agencies stated that attackers succeeded only where basic credential and configuration hygiene was absent, including hardcoded credentials, keys left in code, unauthenticated interfaces, and unpatched software.

Why it matters

  • ·Regulated industries face direct exposure. The FBI cited attacks on banks, healthcare, energy providers, and government operators. It linked ARTEX to the breach of South Korean financial institutions that exposed personal data for approximately 67,000 people. Organizations in these sectors running any of the named AI platforms without enforced credential controls should treat this as an active risk, not a theoretical one.
  • ·The KYC deepfake campaign tests a compliance control that financial and crypto-sector firms rely on for anti-money laundering and customer due diligence obligations. The specific deepfake attempts were rejected by server-side checks in the described campaigns. However, the agents demonstrated near-sufficient frame rate and facial positioning accuracy. Identity verification programs dependent on client-side liveness detection alone face a narrowing margin.
  • ·The continued availability of ARTEX through forks and a commercial version, combined with 1,465 publicly reachable consoles, means the tool's removal from GitHub has not contained the risk. Procurement and vendor risk teams must verify that AI infrastructure components do not expose unauthenticated interfaces. This includes third-party gateways and proxy platforms. Teams must also confirm that credentials are not stored in code or configuration files.

Stay ahead of stories like this

Get every Global AI governance development like this one, plus the rest of the week's developments. Every Thursday.

Free every Thursday. Unsubscribe anytime.

Governance controls affected

What to do now

  • ☐Audit all deployed AI gateway, proxy, and application builder platforms, including LiteLLM, Dify, Langflow, Ollama, One API, New API, Veloera, Sub2API, FastGPT, RAGFlow, Open WebUI, and LobeChat, to confirm no API keys or access credentials are stored in code, configuration files, or publicly accessible artifacts.
  • ☐Confirm that AI platform management interfaces and inference endpoints require authentication before access, and that none are exposed directly to the public internet without a zero-trust or private network control in front of them.
  • ☐Ask your identity and fraud team whether your identity verification checks for onboarding or high-value transactions rely solely on client-side liveness detection, and confirm that server-side verdict validation is in place for all checks.
  • ☐Check whether your organization is running any ARTEX-related tooling or exposed consoles by searching your asset inventory and asking your infrastructure team to scan for the 1,465-category exposure pattern Censys identified as of November 4.
  • ☐Review vendor contracts and security requirements for all third-party AI platform suppliers to confirm they include obligations to avoid default or hardcoded credentials and to notify you of any breach involving your API keys or model access tokens.

What to watch next

The FBI's Flash alert urged all organizations running internet-exposed applications, cloud infrastructure, and partner or supplier systems to implement the mitigation measures it outlined. Compliance teams should monitor sector regulators in financial services, healthcare, and energy. They should watch for follow-on guidance referencing the ARTEX campaigns. The FBI's citation of the South Korean financial institution breach is particularly relevant. The continued spread of ARTEX forks and a commercial version suggests enforcement attention will grow. Teams in jurisdictions with active AI security guidance should map these alerts against their existing control frameworks. This includes jurisdictions covered by the Five Eyes Guidance on the Careful Adoption of Agentic AI Services. The nine AI platform breaches in October and the LiteLLM supply chain attack indicate a pattern that regulators are likely to reference in upcoming supervisory communications.

Related Coverage

Enforcement2026-11-15

DOJ's $400M North Korean Deepfake Bust Breaks Video-Call Verification

The U.S. Department of Justice seized infrastructure behind 'FaceStealer,' a North Korean social engineering scheme the FBI links to more than $400 million in cryptocurrency theft over 18 months. Attackers used compromised accounts, secretly recorded webcam footage, and AI-generated deepfakes of known cryptocurrency figures in fake video meetings. The campaign targeted cryptocurrency and Web3 organizations across multiple countries, with open-source software maintainers also named as targets.

Enforcement2026-11-17

IBM MCP Gateway Flaw Sends User Identity to Servers Configured to Block It

IBM disclosed CVE-2026-110278 in its ContextForge MCP Gateway, versions 1.0.0 through 1.1.0. A flaw causes the gateway to ignore per-server settings that control which upstream MCP servers receive user identity information, such as user IDs, email addresses, and group memberships. IBM recommends immediate upgrade to version 1.1.1, or globally disabling identity propagation as a temporary workaround.

Corporate Policy2026-11-13

United Airlines Blocks AI Agents Over Credential Risk and Federal Aviation Rules

United Airlines announced it will block AI agents and autonomous assistants from booking flights or managing reservations on customers' behalf. The airline cited credential-sharing risks and potential violations of federal aviation rules, including requirements that a human acknowledge banned substances at check-in. United is updating its booking, ticketing, and distribution policies to cover both its own digital platforms and indirect access through travel agencies and aggregators.