← /editions
#157

The FBI Just Called Agentic Hacking Mainstream

// The FBI warns that ARTEX, an open-source multi-agent pentest framework, is being used to reach privileged access in under three hours. The same week, United Airlines banned unauthorized AI agents, the Linux Foundation launched Agent Name Service, and AWS shipped an emulator written entirely by agents.

$ today_in_10_seconds

The FBI now treats autonomous attack agents as a mainstream threat rather than a curiosity. Its November 13 FLASH profiles ARTEX, an open-source agentic pentest framework, and some of the intrusions it describes reached privileged access in under three hours. Elsewhere this week, the industry spent its time deciding which agents to trust.

$ lead_story

The FBI Says Agentic Hacking Has Gone Mainstream

ARTEX is an open-source multi-agent pentest framework written in Go that converts plain-language goals into commands its AI agents execute. The FBI says various malicious actors have used it since at least August 2026 against targets in finance, government, defense, semiconductors, energy, healthcare, telecom, and other sectors. It brings no new exploits. It ships without its own model and connects to external ones, and attacker deployments mostly use cheap Chinese models with few visible guardrails. The developer removed the GitHub repo on October 8, but forks, archives, and a paid ARTEX PRO keep it alive, and researchers counted over 1,600 exposed consoles in early November.

// why_it_matters: The orchestration loop is now the weapon. Your detection window is measured in minutes, and a takedown can't recall an open-source agent.

read more →

$ also_worth_attention

  • United Airlines bars unauthorized AI agents from booking

    Starting November 11, unauthorized third-party AI agents can't book, service, or buy travel through United's digital platforms or indirect channels, and agencies need United's approval to use their own third-party AI tools. United cites privacy, unintended purchases, and federal rules, though agents can still display fares, and CEO Kirby has said he doesn't want another middleman. If you're building consumer agents, treat merchant permission as a first-class dependency. It won't stay a scraping edge case.

    link →
  • Linux Foundation takes on Agent Name Service

    ANS, an open standard that uses DNS infrastructure to provide identity, verification, and discovery for AI agents, launched as a Linux Foundation project at KubeCon NA. When an agent's name is resolved, ANS returns signed records that can be checked against PKI before any interaction happens. That's exactly what merchants need to tell a sanctioned agent from a rogue one. The open question is whether DNS's old, registrar-run trust hierarchy is robust enough for agents.

    link →
  • AWS ships an emulator no human wrote or reviewed

    AWS Emulator covers more than 200 services across 438 API models, and agents wrote and tested all of its code. Three agent teams split the work: one maintains tests, one hunts issues, and one fixes them, because agents succeed more often on bounded tasks. The practical win is that coding agents can test against AWS without being handed real AWS credentials, which is a meaningful property after the lead story.

    link →

$ bigger_picture

The trend behind the headlines

Look at the same week from three sides. Attackers run agent swarms against your identity layer. Merchants answer by banning agents outright. Standards bodies race to give agents names you can verify. AWS's emulator is the builder's version of the same idea: well-specified tasks, automated verification, and no real credentials in the loop. The common thread is that agent capability is no longer the constraint. Trust is, meaning who the agent is, what it's allowed to touch, and how fast you notice when it's lying.

$ tool_of_the_day

FBI FLASH-20261113-001 (primary source)

Skip the vendor takes and read the advisory itself. The FBI says it covers ARTEX's TTPs and indicators of compromise, plus mitigation steps against ARTEX and similar agentic frameworks. If you run agents with cloud or AD credentials, use it as a threat model for your own stack too.

explore FBI FLASH-20261113-001 (primary source) →

$ subscribe --free

Get the next edition in your inbox

// DAIly Agent covers agentic AI three times a week

// free, no spam, exit anytime