OPENAI
How OpenAI built a digital fortress around ChatGPT
OpenAI has tightened its internal security as concerns grow over AI tech theft, particularly from rivals in China.
The company has been rolling out stricter access rules, more thorough staff vetting, and offline setups to protect its most sensitive work.
Things sped up after Chinese start-up DeepSeek launched a competing model in January, which OpenAI believes was built using its own tech through a method known as “distillation”.
Now, only selected staff are looped into certain projects, in a process called “tenting”.
Even conversations in the office are kept hush-hush unless you’re part of the project.
On top of that, systems are kept offline, biometric scans control who enters which rooms, and default settings block internet access unless explicitly approved.
In short:
OpenAI’s tightened its grip on internal access after fears of AI model copying.
Staff now work under stricter rules, offline systems, and tighter physical security.
The broader backdrop? Rising US–China tech tensions and a push to safeguard innovation.
Secrets don’t leave Slack
This shift comes as the US continues to crack down on foreign access to advanced tech, especially in AI.
OpenAI has brought in major security hires, including a former Palantir executive and a retired army general, to lead these efforts.
The company says it’s investing heavily in safety and privacy, aiming to set the standard across the industry, not reacting to any single event.
What happens in OpenAI, stays in OpenAI.