White hat who breached OpenAI says frontier labs' security is alarmingly weak
Miles_Brundage · x · 2026-09-20
s1r1us, one of the white hat hackers who recently breached OpenAI, clarified that finding one serious exploit doesn't mean knowing how to secure such a complex organization, and that OpenAI has far more experienced security staff. He confirmed current models already have serious offensive cyber capabilities. Nathan Calvin added that frontier labs claim world-historic, potentially catastrophic stakes yet run security practices comparable to a random game or B2B SaaS company—sometimes worse—and that employees' concerns appear sincere and increasingly justified.
More from Models
- Free DiffusionGemma endpoint returns sentence probabilities in ~230ms, OpenAI-compatible — victormustar · 2026-09-21
- grok 4.7 spotted in OpenCode; Opus 5.5, GPT-6 Sol and Kimi K3.1 rumored next — haider1 · 2026-09-21
- Gewell: custom Gemma 4 inference engine cuts KV cache VRAM to 0.625x, losslessly — stoppableDissolution · 2026-09-21
- $40M Bet on Calibrated Confidence, Yet TypeSafe Publishes No Calibration Data — prakersh · 2026-09-21
- Dev urges PAW to just say it's a LoRA compiler for sub-1B models instead of hiding the mechanism — willcb · 2026-09-21
- Hemmingway-1: Qwen3.8-27B finetune that writes like a person — paf1138 · 2026-09-21