Model3h ago
Model
OpenAI Updates Model Cybersecurity Safeguards
OpenAI implemented new testing protocols following recent third-party cybersecurity evaluation incidents. These safeguards aim to harden model defenses against adversarial attacks. The lab focuses on refining how external testers interact with GPT-4. This update provides a more controlled environment for vulnerability research without compromising the stability of production AI models.