Breakthrough in Ransomware: ESET Unveils AI-Driven PromptLock
Cybersecurity experts at ESET have unveiled an innovative ransomware variant, dubbed PromptLock, marking the inaugural instance of malware employing generative artificial intelligence (GenAI) for executing attacks.
The discovery, attributed to senior malware analyst Anton Cherepanov and his associate Peter Strýček, reveals that PromptLock harnesses a locally hosted AI language model to create pernicious Lua scripts instantaneously—a transformative leap in the arsenal of cybercriminals.
Diverging from conventional ransomware, which typically depends on static codebases, PromptLock autonomously navigates local filesystems, scrutinizing data content to determine whether to exfiltrate or encrypt files, leveraging pre-formulated text prompts.
Crafted in Golang and employing the SPECK 128-bit encryption algorithm, this malware exhibits cross-platform adaptability, engaging with Windows, Linux, and macOS systems. Notably, an embedded data destruction function remains dormant, suggesting PromptLock currently serves as a proof-of-concept (PoC) rather than a fully-fledged menace.
Cherepanov remarked, “The advent of innovations like PromptLock signifies a substantial shift in the cyber threat landscape. The accessibility of AI technology streamlines the initiation of sophisticated attacks, diminishing the necessity for teams of adept developers. A well-tuned AI model can now engender intricate, self-evolving malware, presenting considerable hurdles for cybersecurity defenders.”
Employing OpenAI’s open-weight gpt-oss:20b model and interfacing locally via the Ollama API, PromptLock generates dynamic scripts without reliance on external servers, rendering it more elusive due to minimal network activity. Moreover, the malware incorporates a Bitcoin wallet address closely associated with Bitcoin’s enigmatic creator, Satoshi Nakamoto, intended for ransom payments; however, no confirmed real-world attacks have emerged as of yet.
ESET’s revelation, published on August 27, 2025, follows several samples uploaded to VirusTotal from the United States, indicating that PromptLock is still in early developmental stages. The institution has classified the malware as Filecoder.PromptLock.A and disseminated technical details to notify the cybersecurity community.
Amplifying concerns surrounding AI-driven cyber threats, a concurrent report from Anthropic indicated that cybercriminals—including a group from the UK identified as GTG-5004—have employed the Claude model to engineer ransomware featuring advanced evasive maneuvers, highlighting the swift incorporation of AI by malicious actors.
Experts caution that PromptLock’s capability to create diverse scripts for each execution complicates detection, rendering traditional indicators of compromise (IoCs) inconsistent and diminishing the efficacy of conventional antivirus solutions.
Nathan Webb, principal consultant at Acumen Cyber, underscored the ramifications of PromptLock’s emergence: “This appears to be the inaugural occasion of AI-enriched ransomware identified in the wild. Its capacity for on-the-fly script generation offers attackers unparalleled flexibility.”
ESET recommends that organizations closely monitor Lua script execution and proxy tunneling related to the Ollama API to mitigate potential threats.
With the accessibility of AI tools on the rise, experts anticipate a proliferation of similar threats. “The escalation of AI-infused malware denotes a new frontier in cybersecurity,” Cherepanov observed.
“By disseminating these insights, we aim to catalyze dialogue and preparedness throughout the industry.” ESET remains vigilant in tracking the evolution of PromptLock, advocating for defenders to adapt to this shifting threat paradigm. (ILKHA)
Source link: Ilkha.com.
