Vibe Coders and AI: the hidden security risk in automatically generated code
The widespread use of AI to generate code is introducing new risks to corporate security. From non-technical employees developing applications without supervision to hidden vulnerabilities in automatically generated code, companies must adopt rigorous measures to mitigate these dangers. Discover how to ensure the security of AI code with specific protocols and advanced tools.
KEY TAKEAWAYS
- Vibe coders and AI generate insecure code: only 11.8% of solutions are protected
- Critical risks: backdoors, ghost dependencies, and leaks of sensitive data
- Defense strategies: static analysis, dependency scanning, and specialist review
- Compliance and security: the importance of rigorous protocols for the integration of ML models
- Technical solutions: SCA, software supply chain security, and automated scanning
Transcript
A young developer, dubbed a "vibe coder," created an unauthorized web application for a European company, exposing sensitive customer data and private contracts to critical risks. The application uses a dangerous combination of public cloud infrastructures and open-source services, hosted on the employee's domain, storing data on Supabase's free plan, and deployed directly from his GitHub repository to Cloudflare's serverless infrastructure. Despite security concerns raised by the IT manager, the CEO downplayed the issue.
"Vibe coders" are developers who prioritize functionality over security. According to the SusVibes benchmark, only 11.8% of solutions generated by unguided agents were secure. The rise of AI code generation tools further increases security risks, as developers become less critical of AI-generated code.
To mitigate these risks, companies are adopting strategies such as treating AI-generated code as untrusted input, running static analysis, secret scans, and dependency checks. Experts emphasize the importance of human judgment in approving AI-assisted changes and using software composition analysis tools to identify vulnerabilities.
The unauthorized use of cloud infrastructures also raises concerns about the company's NIS2 compliance and cyber insurance, potentially invalidating policies due to negligence. This situation highlights the need for increased investment in security budgets to keep pace with software development and AI adoption.