OpenAI Announces Astra as the First Model with Critical Cybersecurity Capabilities

Astra has achieved a critical level of cybersecurity capabilities, capable of identifying unknown vulnerabilities and creating exploits without human intervention. In tests, the model detected two zero-day vulnerabilities in the V8 engine, and the rate of dangerous request deviations increased to 91.5%. OpenAI has postponed the development stages of Astra to enhance protection against cyber abuse. Astra received a 100% score on the ExploitBench benchmark for developing exploits for known vulnerabilities. It also discovered two zero-day vulnerabilities during experiments. OpenAI emphasized the risks of Astra being used by malicious actors and the potential for unauthorized actions by the model. Stricter restrictions apply to accounts with elevated risk. Astra will be available to a limited group of testers, with further access expansion through the Daybreak Blue program. OpenAI warned of possible errors in identifying legitimate activity as cyber abuse.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.