Amid mounting fears of foreign espionage and alleged model theft by Chinese rival DeepSeek, ChatGPT-parent OpenAI has reportedly ramped up its internal security protocols.
What Happened: OpenAI has tightened security to prevent corporate espionage and leaks of its foundational model technologies, reported the Financial Times, citing several people close to the company.
This follows allegations that Chinese AI startup DeepSeek used a technique known as “distillation” to copy OpenAI’s models and build a rival system.
The company has adopted an internal “deny-by-default” egress policy, cutting off internet access to internal systems unless explicitly permitted, the report said.
Also Read: Sarepta Therapeutics Stock Plummets Over 40% On Second Fatal Case Linked To Its Gene Therapy
It has also installed biometric fingerprint scans at secure facilities and restricted employee access under a system known as information “tenting.”
For instance, staff working on the internally code-named “Strawberry” model were warned not to discuss the project outside the designated “tent,” ensuring only authorized personnel had access to critical conversations and code, the report noted.
“It got very tight — you either had everything or nothing,” a source told the publication, describing the new protocols. Over time, more staff are being granted limited access to specific components without visibility into broader projects.
To lead this transformation, OpenAI hired Dane Stuckey, former chief information security officer at Palantir Technologies, Inc. PLTR, last October. He now oversees cyber and data defense alongside Matt Knight, OpenAI’s VP of security products.
The company has also brought on retired U.S. Army General Paul Nakasone to its board, bolstering its defense posture.
Why It’s Important: As tensions escalate between the U.S. and China over AI leadership, OpenAI’s moves reflect a broader concern that foundational model data could be weaponized by geopolitical adversaries.
U.S. authorities have warned that China is aggressively targeting American tech firms to gain access to advanced models and IP.
OpenAI insists these measures are part of its commitment to security and not a response to a specific breach.
Last month, OpenAI also landed a $200 million defense contract to build artificial intelligence tools for national security purposes.
Read Next:
Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors.
Photo courtesy: Svet foto / Shutterstock.com