OpenAI Shifted 1 in 4 Production Engineers to Security: Why AI Engineers Need Cybersecurity Skills

OpenAI President Greg Brockman said the company temporarily redirected 25% of its production engineers toward defensive security work, putting their existing projects on hold while the company investigated and addressed security weaknesses.

The move highlights an important change in AI engineering.

Security is moving inside the build process

As AI systems become more capable and connected to external tools, browsers, code and infrastructure, security increasingly becomes part of building AI systems rather than something considered only after deployment.

For engineers, understanding permissions, sandboxing, infrastructure security, monitoring and secure system design could become increasingly important alongside traditional AI skills.

Context

This followed the disclosure covered in Three Researchers + AI Breached OpenAI.