Forks and copies of the ARTEX AI penetration-testing framework remain publicly available and capable of executing credential attacks and cloud breaches at machine speed across multiple industries, even after the original developer made the project closed-source1.
Check Point Research documented criminals using ARTEX to automate attacks against multiple services, with the framework coordinating multiple AI agents to plan security tests, conduct reconnaissance, test vulnerabilities, and document findings with limited human input2. Existing copies, backups, and maintained forks of ARTEX remain publicly accessible despite the closure of the original repository.
Three campaigns in detail
Check Point researchers documented three campaigns using ARTEX against iOS app-signing providers, cryptocurrency-related services, and gaming-account marketplaces.
In the first campaign, attackers targeted underground iOS app-signing services that provide certificates for installing apps outside Apple's official App Store. According to Check Point, agents obtained more than 300 certificates and associated passwords, three of which passed the validation checklist described in the report. One unauthenticated API exposed AWS credentials granting write access to an S3 bucket containing iOS app files, manifests, and landing pages. Agents successfully modified distribution assets, though the bucket owner later restored the original manifest.
A third campaign targeted 292 cryptocurrency-related domains. The agents found exposed databases, vulnerable APIs, and personal information. The campaign apparently sought to steal USDT and alter payment configurations. Some agent-reported cashout successes were hallucinations based on API responses indicating success, and Check Point observed no funds reaching the attacker's wallet.
Another campaign targeted 13 gaming-account marketplaces, where ARTEX agents found exposed files and poorly secured administrative interfaces. Attackers obtained more than 190,000 Honor of Kings credentials. Sampling indicated that roughly 3.3% of those credentials remained usable, and the usable total suggested about 6,000 accounts, though that figure was an extrapolation rather than an individually verified count.
Model-switching and safeguards
Check Point observed an operator switching models after encountering Opus 5.5 safeguards. Chinese models, particularly the DeepSeek Flash family, predominated in the 10 analyzed instances.
Dark Reading reported that 45 similar pen-testing tools alongside ARTEX forks remain capable of executing credential attacks and cloud breaches at machine speed.
ANALYSIS The persistence of publicly available forks, combined with the low barrier to swapping in less-guarded language models, means the attack surface ARTEX opened has not contracted despite the original project going closed-source.
Vector Wire previously covered Storm-0940's use of the ARTEX framework against ConnectWise's supply chain, during which ConnectWise ran around-the-clock tutorial sessions to help managed service providers check for indicators of compromise[1].