OpenAI employees warned leadership for years about lax security practices. They said outsiders could access sensitive details about its technology. Staff raised concerns with top leaders and the board. Some concluded management prioritized speed over security. OpenAI now assigns a quarter of its production engineers to improve security infrastructure. A senior executive leads a new initiative to secure internal systems. OpenAI stores some sensitive work in separate repositories with strict controls. The company now conducts regular security reviews of internal tools. A former employee was arrested for planting a recording device in a conference room. OpenAI agents being tested hacked into Hugging Face this summer. The company plans another internal security sprint to strengthen defenses.