FBI warns ARTEX AI tool accelerates cyberattacks | Magica
FBI warns ARTEX accelerates attacks, but speed does not ensure a breach
Editorial Team
••📖8 min read
The FBI warns that ARTEX can accelerate intrusions, with some reaching privileged access in under three hours. Microsoft also documents unsuccessful mass targeting, while compromised backup-service credentials opened a separate route into customer environments.
The FBI says attackers have used ARTEX globally since at least August 2026, with some intrusions reaching privileged access in under three hours.
Microsoft observed more than 151,000 failed logons in a password-spray operation without a successful authentication; compromised application credentials created a separate route into customer environments.
Removing its original repository has not stopped distribution; exposed-console counts leave the scale of malicious use unresolved.
Attackers are using ARTEX, an open-source framework originally intended for authorized penetration testing, to compress reconnaissance, exploitation and data theft into automated operations. In a November 13 FLASH advisory, the Federal Bureau of Investigation warned that some intrusions had progressed from initial targeting to privileged access in under three hours.
The warning concerns the time available to contain an intrusion, not an ability to breach every target. Microsoft, whose threat-intelligence investigation supplied the advisory’s under-three-hour examples, says most targeting in the campaign it examined was unsuccessful. Where attackers did obtain privileged access, noisy activity could still outpace a coordinated response.
The bureau's warning draws together campaigns affecting financial services, government, technology, defense, healthcare and other sectors. Its evidence points to automation of familiar weaknesses—exposed applications, stolen credentials and excessive permissions—rather than a requirement for an entirely new class of vulnerability.
ARTEX translates an operator's plain-language objective into commands executed by specialized agents. Written in Go, it supplies the orchestration rather than its own AI model: the FBI lists integrations with Claude, GPT, DeepSeek, GLM, Grok and Kimi.
Its developer, identified as Chinese cybersecurity engineer Li Puhua in a report citing The Wall Street Journal, uses the account Autumn-27. The developer also created ScopeSentry, an asset-discovery and vulnerability-scanning platform. Censys describes ARTEX as an AI agent layer built on that reconnaissance foundation: discovered domains, addresses and endpoints feed its asset graph, according to its .
Jensen Huang has reportedly discussed returning to Nvidia three days after resigning. Supporters want reinstatement during an investigation, while directors have installed interim leadership and begun a permanent CEO search.
The FBI dates ARTEX's original release to July 2026, consistent with Censys’s July 26 date; the report identifying its developer instead gives August. The developer denied involvement in the attacks, said the tool was intended for testing authorized assets and made the repository unavailable on October 8 after reports of abuse. Community forks, localized variants, archived copies and a commercial edition called ARTEX PRO remain available, the bureau says. Censys also found rebrands such as ScopeWeaver circulating after the takedown.
Model access is another dependency. The FBI says DeepSeek Flash models predominated in analyzed attacker-controlled instances, citing capability, relatively low cost and a lack of observable guardrails when requests were framed as penetration testing. That is an observation about those instances, not a benchmark of every model's safety or effectiveness.
There is also evidence of attackers shifting operating costs onto victims. Amazon's threat-intelligence team described Chinese-nexus groups stealing AI access credentials and using the stolen keys to fund inference for further credential hunting, according to a report on its findings. The available evidence does not establish a comparable dollar cost per successful intrusion.
The same report says Amazon identified another campaign targeting cryptocurrency exchanges with real-time deepfake injection to bypass know-your-customer verification. That account broadens the reported uses beyond conventional network intrusion; it does not establish how often those bypass attempts succeeded.
Company-reported Microsoft telemetry recorded 151,000+ failed logons, peaking at 2,900 failures per minute. No successful authentication was observed in this password-spray operation; downstream application access was separate. Source: Microsoft Security.
Failed password attacks preceded abuse of trusted applications
The ConnectWise Cloud Backup campaign shows why the risk extends beyond an organization's own public website. Attackers compromised application identities used by the backup service, formerly SkyKick Cloud Backup, then used permissions already granted by Microsoft 365 customers to authenticate into their environments, a detailed account of the campaigns explains.
Microsoft attributes that operation to Storm-0940, a China-state-aligned actor active since 2021 whose record includes data theft, password spraying and exploitation of internet-facing applications. In its October investigation, Microsoft says previously authorized applications let the attackers operate without an interactive user sign-in. It says the activity did not result from a vulnerability in, or compromise of, Microsoft's corporate systems, products or services.
The preceding password attacks show the limits of scale. Between October 10 and 14, Microsoft observed a password-spray campaign—trying passwords across accounts—from Storm-0940 infrastructure targeting 327 accounts across eight organizations. It recorded more than 151,000 failed logons, peaking at 2,900 failures per minute on October 11; lockout responses accounted for nearly 98% of the activity. No successful authentication, token issuance, Microsoft Graph call or directory change was observed in that operation. Those figures describe the password-spray campaign, not the success rate of the separate downstream application-access activity.
After gaining control of the backup application's identities, attackers progressed in some customer environments to reconnaissance, application credential modifications and changes involving identities, groups and privileged roles. Microsoft warns that newly added credentials could preserve a backdoor even after the original compromised application was remediated.
The scale needs qualification. The FBI describes the ConnectWise compromise as enabling access to more than 2,250 downstream tenants; Microsoft describes attempts to authenticate against more than 2,250. The campaign account cites security firm Huntress as identifying at least 162 compromised tenants at the time. These are different measures. Microsoft also cautions that an affected application or successful authentication does not necessarily mean customer data was accessed or exfiltrated.
Other campaigns establish that the consequences are more than hypothetical. The FBI cites theft of personal information relating to approximately 67,000 people in attacks on South Korean financial institutions. The campaign account also describes South Korean police arresting seven suspects linked to intrusions at more than 90 organizations and theft of approximately 8.2 million records, with ARTEX reportedly used in most intrusions.
Separately, CrowdStrike reported that the cyberespionage group it tracks as Vapor Panda used ARTEX against more than 200 organizations across 14 countries, compromising at least nine, according to the same account. Target counts, confirmed compromises and stolen records describe different aspects of harm and should not be added together.
Company-reported Censys detections of tools sharing hosts with ARTEX. Hosts can carry multiple detections; co-location does not establish which tool performed an intrusion. Source: Censys.
Exposed consoles are an incomplete measure
The FBI cites more than 1,600 publicly exposed ARTEX consoles as of early November, attributing the figure to Censys. But Censys, a cybersecurity firm that detects exposed consoles through internet scans, gives 1,465 hosts in its analysis’s summary, dated as of November 4; the body dates the same count to November 3. About 650 were in two blocks of identically configured hosting, leaving roughly 815 distinct-looking deployments. The publications do not explain the numerical difference.
Censys also found that 155 hosts, or 10.6% of its observed total, ran another detected threat or offensive tool. That is evidence of overlap, not proof that every ARTEX installation is malicious: Censys says some likely belong to authorized testers. Conversely, consoles need not be public, so the survey cannot count deployments hidden on private networks.
ARTEX is part of a wider toolset. Censys found another AI penetration-testing framework, CyberStrikeAI, on 70 of the same hosts, and also observed PentAGI, HexStrike AI and RedAmon alongside ARTEX. Conventional command-and-control and tunneling tools appeared too. Co-location shows operators combining tools; it does not establish which agent performed a particular intrusion or that ARTEX uniquely enables these capabilities.
Operators also expose their own evidence. Censys found 32 open directories on ARTEX hosts containing material such as post-exploitation tools and implants. It had seen the exposed directory on the South Korean campaign’s server from August 18, seven weeks before CrowdStrike published its findings.
The numbers support concern about availability. They do not establish the number of attackers, victims or successful intrusions, or isolate how much faster ARTEX is than a comparable operation without it. The under-three-hour cases demonstrate that rapid compromise occurs; they are not an average attack time.
The next decision is whether trusted access is still safe
The FBI's recommendations concentrate on reducing the opportunities that automation can exploit: inventory peripheral internet-facing systems, restrict administrative interfaces, apply least privilege to application permissions and put rate limits and abuse controls around sensitive APIs.
For organizations with affected ConnectWise integrations, the bureau recommends reviewing even previously authorized applications, searching by application ID, examining service-principal sign-ins and credential changes, and keeping affected connections disabled until formal restoration instructions arrive. It also urges preservation of evidence without delaying urgent containment, and advises against deleting enterprise applications as the first response. Microsoft says responders must verify that credential rotation and tenant remediation removed remaining unauthorized access, while assessing the effect on backup and restore workflows.
The advisory cautions against treating historical IP addresses as conclusive: cloud infrastructure can be reassigned, and indicators need corroboration with current telemetry. Its own observations include lateral-movement attempts blocked by endpoint protection, a reminder that automation does not guarantee success.
A report on defensive tools says cybersecurity researcher MalwareMustDie also released an open-source ARTEX detection pack and playbook, covering network requests, identity abuse, cloud activity and automated attack patterns. That playbook likewise warns that infrastructure indicators can age or produce false positives.
What remains missing is a consistent measure of outcomes: how many attempted intrusions succeed, how much time automation saves against comparable targets, and how quickly defenders contain them. The FBI asks incident reporters for affected accounts, data accessed or stolen, time to detection and time to containment. For affected backup customers, the immediate question is whether disabling the compromised connection also removed the access established afterward. The answer requires evidence of credential, permission and resource activity, beyond blocking the original application.
House China committee chairman John Moolenaar demanded answers from Commerce Secretary Howard Lutnick over delayed restrictions on Nvidia customer Megaspeed. The dispute concerns enforcement, overseas access to advanced chips and commercial influence as President Donald Trump prepares to visit China.
Apple reportedly still aims to unveil iPhone-linked glasses at WWDC 2027 and ship before year-end. Siri control of existing apps is the software bet, but delays cloud the schedule and Meta already has an established eyewear lineup.
Treasury Secretary Scott Bessent urged Nvidia to recruit an outside CEO and explain within 10 days whether suspect chip sales inflated earnings. The board is considering internal and external candidates, while export enforcement also faces scrutiny.
Congressional scrutiny and company investigations are testing what Nvidia knew about Megaspeed, a Singapore cloud operator whose Malaysian subsidiary bought nearly $2 billion of its technology. Megaspeed denies wrongdoing; the available reporting has not conclusively traced suspected chip diversion to China.
Commerce Secretary Howard Lutnick is backing narrower restrictions on Chinese access to overseas AI computing, according to a report examining his former firm’s financing relationships. Bitdeer filings establish Cantor’s role and the scale of its client’s planned expansion, but do not establish improper influence on policy.
Microsoft said services were fully restored on Saturday after a global Microsoft 365 outage disrupted email, healthcare communications and public services. It blamed an earlier network capacity upgrade exposed by a later change, with a full technical review still to come.
Cursor says a key exposed by Fireworks led an attacker to unredacted credentials in historical coding-agent records and then its extension cache. Up to 5,000 users potentially received malicious Claude Code builds. Cursor has moved inference off Fireworks; successful endpoint compromises remain uncounted.
Police are investigating a top tax-accountant exam scorer's admission of AI-assisted cheating, while the government reviews national qualification tests held in 2026 and prepares stronger controls.
An early tally puts Japanese listed companies’ first-half net profit growth at 43.2%, led by manufacturers. Filings show AI component demand alongside currency benefits and non-AI earnings gains, while procurement delays and canceled infrastructure reservations expose the costs of expansion.
TSMC plans to back promising chip startups through a new investment arm, alongside separate commitments to OpenAI and Anthropic. The strategy reaches toward future customers and the AI developers shaping chip demand, but the investment rights remain unclear.
Nidus has raised $30 million across pre-seed and seed rounds to automate wire processing in existing U.S. defense factories. Its founders bring manufacturing and government-deployment experience, but customer access and proprietary data still need to translate into measurable production gains.
Nebuly is funding US and European expansion for its AI business-value platform. Adjustable assumptions and private deployment are part of its pitch, but rivals offer similar measurement tools and verified customer returns remain undisclosed.
Han Won-seok says he used a miniature camera, generative AI and a smartphone in tax accountant and CPA exams. His confession puts scrutiny on exam security after proctors reported no suspicious behavior.
Reports linked to AI and child sexual exploitation rose from about 4,700 in 2023 to more than 400,000 in 2025, Childlight finds. The increase exposes growing threats and weaknesses in reporting systems, but does not measure unique victims or confirmed AI-generated images.
Megaspeed says Alice Huang left on August 30, 2025, and has retained outside lawyers for an internal compliance investigation. Singapore’s inquiry began months before that departure; authorities later found insufficient evidence that Huang had committed a crime, while the investigation into the company continues.
HP’s US listing puts a $132,000 starting price on its ZGX Fury shared AI system. Its 748 GB of coherent memory supports a local-computing pitch, but competing GB300 machines, dedicated power requirements and workload-specific operating costs shape the buying decision.
Apple's new U.S. study will examine how up to 200,000 adults understand and use health answers from OpenEvidence, a clinician-focused AI company. Optional personal health data adds context, but competing chatbots already offer data connections, and the study has yet to demonstrate better health decisions.
A $1.5 trillion valuation underpins the estimate of 4,250 Anthropic employee millionaires. The offering remains unpriced, selling restrictions limit liquidity, and the latest filing shows computing costs complicating the company’s rapid growth.
Supermicro says Rubin racks are shipping and Nvidia claims production at scale, while supplier reporting points to larger deliveries in 2027. Memory constraints and differing forecast scopes leave the pace of the transition unresolved.