New Study Identifies 53 Slopsquatting Targets Across 5 Frontier LLMs
Study identifies 53 available package names that LLMs hallucinate across PyPI and npm.
Summary
Independent research analyzing nearly 200,000 code-generation responses from five frontier LLMs (Claude, GPT, Gemini, DeepSeek) found that all models generate the same nonexistent package names at rates of 4.62–6.10%. Security reviews by PyPI and Socket identified 53 registrable names (41 on PyPI, 12 on npm) that could be exploited via slopsquatting—where attackers register hallucinated package names and wait for AI tools to recommend them to developers. A single malicious registration could target users across multiple model providers, creating a supply-chain risk for AI-assisted code generation.
Full text
Security NewsWhite House Launches Gold Eagle Initiative to Manage Surge in AI-Discovered VulnerabilitiesThe White House’s Gold Eagle Initiative aims to coordinate AI-discovered vulnerabilities, validate findings, and accelerate patching across critical software.By Sarah Gooding - Jul 17, 2026