OpenAI has revamped its security to block corporate espionage. The Financial Times reports OpenAI sped up this clampdown after Chinese startup DeepSeek dropped a competing AI model in January. OpenAI claims DeepSeek used “distillation” to copy its tech.
The new measures include strict “information tenting” to limit who can access sensitive algorithms. During the o1 model build, only cleared team members could even talk about it in shared spaces.
OpenAI also moved key tech to offline systems. Biometric fingerprint scanners guard office entry. The company enforces a “deny-by-default” internet policy requiring explicit permission for outbound connections. They’ve beefed up data center security and added more cybersecurity staff. 
The changes reflect worries about foreign spies stealing OpenAI’s IP. But with AI poaching wars and frequent leaks of CEO Sam Altman’s remarks, internal security may be a target too.
We’ve reached out to OpenAI for comment.
