OpenAI has significantly revamped its security operations to shield its intellectual property from corporate espionage, particularly in response to claims that Chinese AI startup DeepSeek copied its models. The Financial Times highlights that OpenAI is enforcing stricter access to sensitive data and enhancing staff vetting. This includes implementing information “tenting,” which limits personnel access to new algorithms and requires fingerprint scanning for entry into secure rooms. Additionally, OpenAI uses a ‘deny-by-default egress policy’ to prevent unauthorized internet connections and has bolstered physical security at its data centers. Microsoft researchers have suggested potential data exfiltration related to DeepSeek, which is positioned as a significant competitor with its open-source R1 reasoning model. OpenAI is also verifying identities with government IDs for developers seeking access to its advanced AI models. This heightened security reflects the growing concerns over the competitive threat posed by Chinese AI technologies in the industry.
Source link
