ANY.RUN linked 1,258 public sandbox analyses to fake AI campaigns between April and October of 2026 — a striking tally that frames a simple, persistent threat: attackers are imitating trusted AI products to seed malware, steal browser sessions, and compromise business accounts.
How attackers impersonate Claude, ChatGPT, Copilot, Gemini, and other AI services
The campaigns investigated by ANY.RUN rely on familiarity. Attackers impersonate Claude, ChatGPT, GitHub Copilot, Gemini and other AI brands, placing lures in sponsored search results, legitimate hosting services, shared AI conversations, and counterfeit download websites. One analyzed Claude impersonation shows the chain: clicking a Windows download button copies a command to the clipboard and displays instructions to run it through PowerShell; the command then launches mshta.exe, retrieves a malicious HTA file, and triggers an encoded PowerShell execution chain.
That sequence creates a difficult detection scenario: the employee initiates the process voluntarily, the attack uses legitimate Windows utilities, and the malicious payload appears only after multiple execution stages. ANY.RUN’s analysis emphasizes how routine software installation flows can be weaponized without obvious, single-point indicators.
Why these campaigns strain MSSP operations
MSSPs supporting U.S. businesses face amplified workload: the source notes U.S. businesses accounted for 16% of country-attributed targets. Similar attacker behavior appearing across different customers can trigger repetitive investigations, consume analyst capacity, and extend response times — especially where customers use different technologies, policies, and risk profiles.
- Monitoring and detection coverage: attackers rotate domains, hosting services, and AI brands, reducing the lifespan of individual indicators.
- Triage and escalation management: evasive, multi-stage chains require deeper investigation and increase escalation volume.
- Threat hunting and detection improvement: individual findings must be correlated and reused to prevent repeated work across clients.
These pressures affect the operational metrics MSSP leaders track: mean time to detect (MTTD), mean time to respond (MTTR), escalation rates, investigation cost, and SLA compliance.
Nobody's watching your logs at 2 AM.
Full SOC coverage without building one. Nubivance deploys and manages Rapid7 InsightIDR and MDR for organizations that need detection and response, not another dashboard.
Get coverageReducing Tier 1 workload with behavioral sandboxing
ANY.RUN’s Interactive Sandbox lets analysts reproduce user actions and observe resulting processes, network activity, and payload behavior in isolation. For MSSPs, that capability makes complex, staged investigations manageable within routine Tier 1 workflows rather than routing ambiguous cases immediately to senior analysts.
ANY.RUN reports 30% fewer Tier 1 to Tier 2 escalations when its sandbox is used, alongside a median threat detection time of 14 seconds. The result, the source argues, is a more even distribution of work across SOC tiers: senior analysts focus on complex incidents while junior analysts resolve a larger share of routine cases, helping maintain SLA performance as alert volumes grow.
Extending detection with threat intelligence and threat hunting
Because landing domains and hosting rotate, MSSPs need continuously updated, validated intelligence. ANY.RUN Threat Intelligence (TI) Feeds extract indicators from real malware and phishing investigations for integration into SIEM, EDR, and other security workflows. According to the source, 30% of unique malicious IOCs discovered by ANY.RUN customers during Sandbox investigations are already available in ANY.RUN TI Feeds.
The source further states that ANY.RUN TI Feeds provide 99% unique IOCs compared with other threat intelligence sources and supply access to 24 times more IOCs per incident. Analysts can use Threat Intelligence Lookup to relate suspicious indicators to historical sightings and campaign infrastructure.
ANY.RUN researchers identified recurring behaviors useful for hunting and detection engineering — suspicious PowerShell installation commands, mshta.exe execution, encoded download instructions, and scheduled tasks launching browsers for session theft — allowing teams to search across managed customers for related activity rather than chasing fresh domains each time.
What this means for MSSPs, affected enterprises, and end users
- MSSPs: Adopt coordinated monitoring, behavioral investigation, and intelligence reuse to reduce repetitive investigations, lower investigation costs, and preserve SLA performance as attack variants proliferate.
- Affected enterprises and procurement leaders: Expect similar attacker patterns to appear across vendors and customers; organizations should align their detection and response expectations with MSSP capabilities and the operational metrics those providers manage.
- End users: Routine actions — clicking a download button, copying and executing an apparently benign command — are a primary vector; staged payloads leveraging legitimate utilities complicate detection at the point of initial interaction.
The operational payoff, as presented, is quantifiable: a 36% higher detection rate and up to 58% more threats identified, 21 minutes faster investigations, and fewer escalations — improvements that, according to the source, allow MSSPs to scale services “without proportionally increasing analyst headcount” while reducing the risk of major client incidents. The record from April through October 2026 shows a threat pattern that is not exotic but persistent: attackers reuse familiar brands and routine workflows, and defenders must reuse intelligence, tooling, and behavioral evidence at scale to keep pace.
