FBI warns AI hacking framework Artex is automating intrusions
The FBI says the open-source tool lets attackers chain hacking steps at machine speed. Cisco Talos and Cloudflare report its forks and copycat frameworks are spreading fast.
Key takeaways
- The FBI issued a flash alert saying the AI framework Artex automates intrusions from reconnaissance to data theft.
- The FBI says some intrusions reached privileged access in under three hours.
- The developer removed the GitHub repository on Oct. 8, but forks and copies keep spreading, the FBI says.
- Security firms have tied Artex to China-nexus espionage and to financially motivated groups.
- Talos and Cloudflare advise MFA, fast patching, secret scanning, rate limits and rapid containment plans.
The FBI issued a flash cybersecurity alert on Nov. 13 about Artex, an open-source AI penetration testing framework that the advisory says cybercriminals and nation-state actors are using to break into organizations worldwide, The Record reports. According to the FBI, the tool can take a large language model and use it to "automate multiple phases of cyber intrusions at machine speed, from reconnaissance through data exfiltration."
"Artex significantly reduces the time required to chain attack techniques, with some intrusions progressing from initial targeting to privileged access in under three hours," the FBI said, as quoted by The Record. The agency added that the framework lets people with little technical skill run complex operations against several targets at once.
What is Artex?
Cisco Talos describes Artex as a framework built for authorized security testing. An operator sets a target and a goal, a planner agent splits the goal into tasks, and worker agents carry them out with security tools, logging assets and findings in a shared graph. It can plug in different AI models and needs little input once a task starts, Talos says. The Record says it can use any AI model, including Anthropic's Claude and OpenAI's GPT as well as open-source or Chinese-origin models such as DeepSeek, GLM, Grok and Kimi.
The framework was published on GitHub under the name Autumn-27. Talos dates the release to July 2026, while The Record says August. In a statement on Oct. 8, the developer said the abuse ran counter to the tool's original intent, Cloudflare reports, and denied responsibility for crimes committed with it, according to The Record. Cloudflare dates the takedown of the repository to the same day; The Record says it was eventually removed. The FBI said the framework still spreads through forks, local variants, archived copies and a commercial version.
Who is using it?
The Record, summarizing the advisory, says Artex has been used in attacks targeting banks and other financial firms, government bodies, tech and chip companies, defense, energy and healthcare organizations, manufacturers, telecom operators and law firms. The FBI said models from Chinese developers are the most common in attacker-run instances it analyzed, with DeepSeek's Flash family models dominating; the advisory attributes that to their "balance of capabilities, relatively low cost, and lack of observable guardrails when framed for penetration testing." None of the cited reports includes a response from DeepSeek.
Talos summarizes campaigns that other security firms have tied to the tool. CrowdStrike attributed an espionage campaign that began Oct. 17 to a China-nexus group it calls VAPOR PANDA, whose Artex operations reached over 200 organizations across at least 14 countries; CrowdStrike confirmed post-compromise activity at nine. Microsoft linked Artex activity to Storm-0940, which it assesses as a "China-state aligned actor," after a compromise of ConnectWise application credentials, and to a financially motivated, Cambodia-based group targeting financial firms mainly in Southeast Asia. Talos says the evidence it examined in one South Korean case supports an association with Artex but does not by itself prove the full framework was used.
How big is the spread?
The FBI advisory cites research by Censys, The Record reports, that found more than 1,600 Artex consoles exposed on the internet, with the real number likely higher. Cloudflare says that in three days it saw more than 50 families of AI penetration testing frameworks try to scan its customers, that daily requests from such tools have grown thirtyfold since Oct. 1, and that Artex and its forks made up more than 60% of the known AI testing orchestrators it tracked as of Nov. 12. It counts more than 12,000 related repositories, several hundred of them forks with lineage from the original.
How can organizations protect themselves from these attacks?
Talos recommends focusing on basics that slow automated attacks:
- Credentials: require multi-factor authentication for remote and privileged access, and add rate limits and alerts for password spraying.
- Exposure: patch quickly, remove default credentials, and scan public code repositories for exposed keys, rotating any that are found.
- Detection: watch logs for bursts of unusual activity, and have plans ready to revoke tokens and isolate systems quickly.
Cloudflare, which sells such services, recommends web application firewalls with rate limits, bot defense and identity-gated access for high-risk applications.
Disclosure: NextFish's AI drafting desk uses Claude, a model made by Anthropic.
Questions readers are asking
What is Artex?
An open-source AI penetration testing framework that uses planner and worker agents to carry out security tasks with little human input. It was built for authorized testing but has been misused, according to Cisco Talos and the FBI.
Who is using Artex in attacks?
Security firms have linked it to a China-nexus espionage campaign, to a group Microsoft assesses as China-state aligned, and to financially motivated criminals. According to The Record's account of the FBI advisory, both cybercriminals and nation-state actors use it.
Can Artex be stopped by taking it off GitHub?
The developer removed it on Oct. 8, but the FBI says forks, local variants, archived copies and a commercial version keep it available.
How can organizations defend against it?
Talos recommends multi-factor authentication, rate limits, prompt patching, removing default credentials, scanning code for leaked keys and preparing to revoke tokens and isolate systems fast.
Sources
- #artex
- #ai-security
- #fbi
- #penetration-testing
- #cyberattacks
Get the day's top AI & tech stories on WhatsApp: Follow the NextFish channel
Get stories like this every morning.
Related
Cybersecurity4 min
ConnectWise urges ScreenConnect client updates as SCKit targets technicians
Microsoft, Huntress and Blackpoint have analysed the toolkit used against MSPs. It is built to jump to technicians, but researchers say they have not seen that work.
Cybersecurity3 min
Landmark Group says Shukran loyalty members' personal data was accessed
Names, contact details and dates of birth were involved, the Dubai retailer says, but not card details, passwords or points. It has not said how many members were affected.
Cybersecurity3 min
Operation Endgame takes down Gootloader, CastleRAT and StealC servers
The Justice Department says Gootloader infected at least 25,000 computers and was used against at least 44 US ransomware victims. Europol counts 235 servers hit in all.
Cybersecurity2 min
TSMC says leaked employee data matches records at a cloud provider
The chipmaker says its own systems show no sign of compromise after a seller offered a 6.9GB file of employee records on hacker forums for three bitcoin.