THE cybersecurity sector is facing an increase in fake AI proof-of-concept (PoC) submissions on public code-sharing platforms, according to a report by VulnCheck. While researchers have reviewed over 17,800 PoCs in 2026, the quality has plummeted, with acceptance rates falling from 51% in previous years to 45% due to many submissions lacking real exploit code. The report highlights 'SloPoCs', which are characterized by excessive emojis and unsubstantiated claims of functionality.
Additionally, repositories containing malware disguised as functional exploits are a growing concern, including instances of malicious modifications in legitimate exploits. Researchers are advised to remain vigilant regarding base64-encoded scripts and overly obfuscated code, as these often bear malicious intent. As the volume of submissions increases, it is essential for security teams to thoroughly vet any code sourced from public repositories.