Security2 publishers3 min readPublished Updated
SilkParasite: AI turns up as a build tool inside a human-run espionage kit, not as the author
Bitdefender says a China-nexus cluster hit Central Asian governments with seven RAT families, five undocumented. The AI fingerprint is in the workflow, not the code.
The Watch · Security desk
What happened
- A previously unreported cyber espionage operation dubbed SilkParasite has been observed targeting government bodies in Central Asia.
- The intrusion set uses seven remote access tool (RAT) families, five of which have never been previously documented: DriveSilkRAT, CookiETagRAT, NomadRAT, GoginRAT, and NodeEdgeRAT.
- SilkParasite was first discovered in late 2025 and is assessed to be a China-nexus threat cluster with medium confidence.
- Bitdefender Labs said: "What makes SilkParasite interesting is the traces of AI-assisted development running through otherwise expert code, which is a different thing from AI-generated malware."
- Unlike operations that rely on AI-generated malware, SilkParasite's arsenal exhibits all hallmarks typically associated with professional espionage tooling developed by a team of human operators, while AI is likely used to streamline the process.
Compiled by The WatchSomething wrong?How this is made
Why it matters
Bitdefender Labs has documented SilkParasite, a previously unreported espionage operation targeting government bodies in Central Asia with seven remote access tool families, five of which have never been publicly described: DriveSilkRAT, CookiETagRAT, NomadRAT, GoginRAT and NodeEdgeRAT [1][2]. What makes the report worth reading is not the tool count but the vendor's reading of provenance: according to Bitdefender, the code shows "traces of AI-assisted development running through otherwise expert code, which is a different thing from AI-generated malware" [4].
That distinction has operational teeth. Bitdefender's assessment is that the arsenal carries all the hallmarks of professional espionage tooling built by a team of human operators, with AI likely used to streamline the process rather than to write the implants [5]. The clearest evidence of the technology, the company says, is a phishing lure that is unmistakably AI-generated, and that lure is also the only place the operators were sloppy, which Bitdefender says raises the possibility the sloppiness was deliberate cover against attribution [6]. Treat generated text in a lure as a hint about tempo, not a verdict on skill.
The rest looks like standard tradecraft executed well. Intrusions begin with password-protected RAR archives carrying malicious Office documents, likely delivered by spear-phishing, with the archive password supplied in the email body; opening the document runs a macro that triggers a DLL sideloading sequence to drop the first stage [12]. Before executing, the macro checks whether Kaspersky's antivirus is installed and running, which Bitdefender reads as an attempt to evade the product's regional prevalence [14]. Lures were tailored to Uzbekistan, Turkmenistan, Kyrgyzstan, Tajikistan and Kazakhstan, several impersonating specific ministries, with one document pulled from a public malware-sharing platform addressed to a Georgian government entity [13].
Almost every tool in the chain uses a plugin-oriented architecture, letting operators add capability on demand, serve payloads selectively per environment and keep the detection footprint small, and upgrade components without replacing the foundations [15]. The seven implants span .NET, C++, Go and JavaScript, with DLL sideloading as the main delivery vector, achieved by bringing their own legitimately signed program and dropping a rogue DLL under a name that executable looks for [16][17]. DriveSilkRAT, written in .NET and C++, uses Google Drive as command and control, polling a specific folder for tasking [18]. Four languages and a bring-your-own-binary pattern is the kind of surface that breaks signature-first detection and rewards behavioural rules on sideload paths.
Attribution rests on reuse, not on the new families. SilkParasite is assessed as China-nexus with medium confidence and was first discovered in late 2025 [3]. Two of the seven families were already known [19]: BLOODALCHEMY, an updated Deed RAT, itself a successor to ShadowPad and thus a descendant of PlugX, both widely used by Chinese groups [8], and an updated SpiceRAT attributed to the Chinese-speaking actor SneakyChef [11]. BLOODALCHEMY was first documented by Elastic Security Labs in October 2023 in REF5961 attacks on governments in Southern and Southeast Asia [9]. Bitdefender notes SilkParasite is the third prominent actor to hit Central Asia in recent years, after UAC-0063 and FamousSparrow [7].
Watch whether the AI-assisted-development signal shows up in other clusters as a repeatable analytic marker, or stays a one-report observation. Watch also whether the five new families reappear elsewhere, since shared plugin frameworks and loader naming will carry attribution weight that generated phishing text cannot.