The new threat created by AI coding tools: Slopsquatting
AI hallucinations are putting software supply chains at risk.
1What is slopsquatting?
Slopsquatting is a novel supply chain attack enabled by AI hallucinations. Coding assistants suggest fake open-source package names to developers. Cybercriminals register these imaginary names to inject malicious code. Unlike typosquatting, it doesn't rely on simple misspellings but uses entirely fabricated yet plausible-looking package names. This emerging threat bypasses traditional security measures in software ecosystems.
2How AI creates the risk
Traditional typosquatting protections blocked misspelled versions of popular libraries. But AI generates original yet fake package names that evade these filters. Models tend to repeatedly produce the same hallucinated package names. Attackers exploit this predictable behavior to target thousands of developers. Package registries aren't yet prepared for this new threat vector emerging from AI-assisted coding workflows.
3Surprising hallucination rates
Research shows LLMs can hallucinate 50-82% of the time when generating code. Even top-performing GPT-4o maintains a 23% error rate. Open-source models carry 4x higher risk. In one study analyzing 576,000 code samples, 19.7% of 2.23 million generated packages were fictional. This represents a critical security vulnerability, especially for organizations using open-source AI coding tools.
4Real-world impacts
Malicious packages uploaded by attackers can remain undetected for months or years. The average lifespan of software vulnerabilities has increased by 85%. AI-suggested fake packages carry names resembling real libraries, earning developers' trust. Single-character differences appear as innocent typos while executing harmful code. This stealth approach makes slopsquatting particularly dangerous for software supply chains.
5Vibe coding amplifies the threat
72% of developers use AI coding tools daily, with over 40% of committed code receiving AI assistance. However, most developers don't verify whether suggested packages actually exist. This 'vibe coding' trend significantly expands the attack surface for slopsquatting. Blind trust in AI recommendations without verification processes creates substantial security risks in modern development workflows.
6Necessary precautions
Organizations must implement additional security layers when using AI coding tools. Package verification processes, dependency scanning tools, and custom security policies need development. Developers should cross-check every AI-suggested package against official registries. Teams using open-source AI models require extra vigilance. This novel threat demands a fundamental rethink of software security approaches in the age of AI-assisted development.