--- language: en title: "Slopsquatting" description: "Slopsquatting exploits names invented by models to register malicious packages. A package’s presence in a repository does not prove that a dependency is trustworthy." canonical: "https://tropk.ai/en/resources/glossary/slopsquatting" last_modified: "2026-10-12T16:09:30.211Z" --- # Slopsquatting Reviewed on 2026-10-07. ## What it means Slopsquatting is the exploitation of names that models invent when suggesting dependencies or other resources. In the documented case of software packages, an attacker can register a name that initially did not exist and add malicious code. Anyone who later follows the suggestion will find a real package, but not the legitimate library they thought they were getting, creating a risk to the software supply chain. ## Why it matters for marketing For marketing, this risk matters when teams use AI assistance to prepare websites, scripts or integrations. A plausible suggestion does not replace technical verification of a dependency. Anyone commissioning the work should require checks of its source and version before execution in systems with access to data. The impact depends on the code installed and its permissions. ## Example in practice Imagine an agency preparing a script to consolidate campaign files. The assistant recommends a library, but the team cannot find any prior documentation for that name. A newly published package appears in the repository. The person in charge halts the installation and checks the maintainer, history and need for the dependency. They choose a reviewed alternative without treating its recent appearance as proof of legitimacy or an attack. ## How to apply or verify Ask the technical lead to check the identity, documentation, source and version of each new dependency. Use review controls and verified versions, keeping a record of choices and avoiding automatic execution of suggestions. Internal repositories and integrity checks help, but require their own maintenance. Research recommends human verification: checking only whether a name exists does not resolve the risk described. ## Important distinctions Package [hallucination](/en/resources/glossary/ai-hallucination) is the suggestion of a nonexistent name; slopsquatting involves exploiting it through a malicious registration. This differs from changing letters in a legitimate name, though it may overlap with other forms of confusion. Neither a new package nor an unfamiliar name proves adversarial intent. The frequency observed in a study depends on the models, tasks and versions evaluated and does not represent a current universal rate. ## Sources and references - [Package Hallucinations: How LLMs Can Invent Vulnerabilities]() — USENIX; Joseph Spracklen e colaboradores ## Continue a leitura - [AI hallucination]() - [Retrieval poisoning]() - [Prompt injection]() ## About the author [Vinicius Pacheco]() Responsible for Tropk’s technical content, focusing on crawling, indexing, structured data, and agent access to content. --- [All pages in Markdown]()