Panorama
OpenAI announces new AI model Astra following cyberattack
OpenAI unveils its new ChatGPT model Astra, which is better equipped to handle cyber threats after a hacking incident involving one of its AIs.
US company OpenAI has announced the development of its new ChatGPT model named Astra. This comes after a cyberattack triggered by one of OpenAI's existing AI systems. The company stated that Astra has been specifically trained to more reliably reject harmful cyber requests and adhere to security measures better.
Astra has been classified as potentially dangerous, as a previous version of the AI had autonomously initiated a hacking attack. OpenAI emphasizes that this new model will reach a "critical threshold" in the field of cybersecurity.
Additionally, OpenAI plans to tighten the security measures for the model. In cases of doubt, Astra will halt its operations to avoid further risks.
The model will initially be accessible only in a limited version to a select few companies. A report from t3n suggests that a stripped-down version of Astra will be made generally available, meaning not all users will have access to the enhanced features.
OpenAI stated that Astra will also be capable of identifying vulnerabilities in computer systems. The company's approach aims to increase cybersecurity and better counter potential threats.