Loading malware…
MalTerminal is an early proof-of-concept LLM-enabled malware prototype, described by SentinelLABS as potentially the earliest known malware to generate malicious payloads at runtime. It was uncovered through hunting for embedded LLM API keys and prompts in malware samples, and was presented by SentinelLABS at LABScon 2024. The tooling included Python scripts and a compiled Windows executable named MalTerminal.exe, with related files such as testAPI.py and TestMal2.py. When executed, it offered an operator a choice between generating ransomware functionality or a reverse shell. It used OpenAI GPT-4 via a hardcoded API key and a deprecated chat completions endpoint to request malicious code generation at runtime. Reported behavior includes generating ransomware, reverse shell, encryption, and exfiltration code in memory without writing the generated payload to disk, complicating traditional static detection. The malware is consistently characterized in the source material as LLM-assisted rather than fully autonomous or agentic, and as an AIM3 Level 1 experimental prototype demonstrating LLM-driven ransomware and RAT concepts. SentinelLABS reported no evidence that MalTerminal or related tools were deployed in the wild, sold, or otherwise operationalized. Detection-relevant artifacts directly mentioned in the content include the hardcoded OpenAI API key usage, embedded prompts, the compiled filename MalTerminal.exe, use of GPT-4, and the deprecated OpenAI chat completions endpoint, which suggests the sample was authored before early November 2023.
Mallory pivots from this family to the IOCs, detections, and named campaigns that touch your stack, and pages you when something new lands.
3 distinct techniques documented for this family, organized by ATT&CK tactic.
15 indicators attributed across vendor reports, sandbox runs, and researcher write-ups. Full values are available in Mallory.
File hashes (MD5, SHA-1, SHA-256) from samples and reports.
11 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
Referenced as an example of earlier LLM-assisted malware where the LLM acts as a component within a human-built workflow rather than as the autonomous operator.
AI-enabled malware that uses OpenAI GPT-4 via a hardcoded API key to generate reverse shells or ransomware at runtime.
AI-powered malware that dynamically generates malicious payloads at runtime by querying a GPT-4 endpoint. It can produce either ransomware functionality or a reverse shell, with encryption and exfiltration code generated in memory to reduce disk artifacts.
LLM-enabled malware that can generate ransomware and reverse-shell functionality (as described in the title/summary).
Match every observed IP, domain, and hash against your live telemetry.
Named campaigns wielding this family, with evidence pinned to each claim.
CVEs this family uses for access and lateral movement.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Every documented technique, ranked by evidence weight.
Reddit, Mastodon, and CTI community discussion around this family.