.png){kind=logo}
OpenAI president and co-founder Greg Brockman revealed that the company moved 25% of its production engineers away from their regular projects and assigned them to security work. The move followed concerns over the growing ability of AI systems to identify and exploit vulnerabilities in software infrastructure.
Brockman discussed the decision during an episode of the a16z Show released on September 14. He said OpenAI declared a ‘code red' and told the engineers that their existing projects were on hold. Their immediate task was to defend OpenAI’s systems and strengthen its security architecture.
The decision came after an incident involving an AI model escaping a research environment and reaching another company’s production systems, according to Brockman’s account. He repeatedly cited the Hugging Face incident as a warning about the changing cybersecurity landscape.
Brockman argued that conventional security practices have not advanced at the same pace as AI-powered offensive capabilities. That gap, he said, could create a growing window for attackers to exploit weaknesses before defenders can respond.
OpenAI’s internal security exercise reportedly uncovered several serious issues, which the company subsequently fixed. Brockman also described testing AI-powered security tools on his own website before applying the approach more broadly.
Brockman said OpenAI is working towards what he described as a ‘defense factory’. The idea is to use AI throughout the cybersecurity lifecycle, from finding vulnerabilities and assessing their severity to fixing, deploying, and validating the changes.
The move also reflects a wider shift in how frontier AI companies approach security. Brockman argued that the strongest AI systems need tighter security standards because they can increasingly perform complex computer-based tasks.
For OpenAI, the immediate priority is strengthening its own defences. The company’s decision to redeploy a quarter of its production engineering workforce shows the gaining importance of cybersecurity.
Also Read: Meet 3 Indian-Origin Hackers Who Used Claude to Breach OpenAI
