Back Feeds.4Sysops

HalluSquatting attack weaponizes AI hallucinations to build botnets

Feeds.4Sysops •IT News • July 8, 2026

Researchers have identified a novel attack vector called HalluSquatting that exploits the tendency of Large Language Models (LLMs) to hallucinate non-existent repository and package names. By predicting which identifiers an AI coding assistant is likely to invent when asked to clone trending software, attackers can pre-register those names on public registries like GitHub or PyPI. These malicious repositories are seeded with instructions that trigger the assistant’s built-in terminal to execute payloads, such as reverse shells or ransomware. Source

The full story

This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.

HalluSquatting Attack Exploits AI Hallucinations to Create Botnets Researchers have identified a new attack vector called HalluSquatting, which exploits AI coding assistants' tendency to hallucinate non-existent software package names. Attackers can pre-register these names on platforms like GitHub and PyPI, creating malicious repositories that trick AI assistants into executing harmful payloads, such as botnet malware. This technique, termed 'adversarial hallucination squatting,' poses a significant risk to systems relying on AI for coding assistance. The research was conducted by a team from Tel Aviv University and Technion, highlighting the potential for large-scale botnet creation. The attack method raises serious concerns about the security of AI-driven software development tools. Current status indicates that the cybersecurity community is on alert for potential exploitation of this vulnerability. 32 sources 65

Extracted Entities

HalluSquatting attack weaponizes AI hallucinations to build botnets

The full story

Extracted Entities

Attack Types (3)

MITRE ATT&CK (2)

Platforms (2)