AI Risk Clock
Doomsday Clock2 min to midnight
→
⚖️ Neutral edit
Forbes2026-05-02

A 22-Year-Old Dropout Just Reverse-Engineered The World’s Scariest AI - Forbes

ModelsResearchSafety

A 22-year-old developer named Kye Gomez has published OpenMythos, a public project that approximates Anthropic's Claude Mythos model. Claude Mythos is described by Forbes as 'the world's scariest AI.' Gomez, a college dropout, reverse-engineered the model based on educated guesses about its core design.

The article notes that the development raises questions about the security of proprietary AI models. If a self-taught developer can recreate a frontier model in days, the 'proprietary moat' around such technology may already be gone. Anthropic has not commented on the project. Gomez claims OpenMythos achieves similar performance to Claude Mythos on several benchmarks, though independent verification is pending.

This event comes amid ongoing debates about AI safety and open-source vs. closed models. Some experts argue that such reverse-engineering efforts could accelerate both beneficial innovation and malicious use. The article suggests that the case of OpenMythos may pressure labs to reconsider their security practices and the public release of model capabilities.

Read this story in another voice
● REC · 2026