OpenAI has ramped up its internal security measures following reports of a potential breach by DeepSeek, which allegedly replicated its AI models using distillation techniques. To safeguard sensitive information, OpenAI implemented stringent access protocols, including fingerprint scans and offline servers, along with a restriction on internet use without prior approval. Only authorized staff can discuss critical projects, such as the AI o1 model, in designated areas. Additionally, OpenAI has increased its cybersecurity workforce and enhanced data center defenses through ‘information tenting’ to protect confidential development details. These security improvements align with OpenAI’s strategic partnership with Oracle, involving a $30 billion deal to lease 4.5 gigawatts of data center capacity in the U.S., essential for supporting its expanding Stargate infrastructure. For those interested in AI, technology, and digital diplomacy, feel free to inquire with our Diplo chatbot for more insights!
Source link
Share
Read more