OpenAI lanza Astra, su modelo más potente y controvertido hasta la fecha

OpenAI has unveiled its most powerful and controversial model to date, GPT-6 Astra, challenging Anthropic's leading position in the AI race. Astra is touted as a general intelligence capable of performing tasks that a human could do, all while being faster and more secure. OpenAI CEO Greg Brockman announced the model's availability to OpenAI customers using Daybreak, their cybersecurity program, with plans to expand to other payment plans, the API, and Amazon Web Services.

The new AI model is designed to work directly within computer programs, capable of tasks such as creating legal contracts, building 3D video games, and designing circuit boards. OpenAI claims Astra is the most intelligent and aligned model in the world, setting a new standard for various applications including cybersecurity, science, and professional work.

The AI has also demonstrated the ability to find and exploit unknown security flaws in systems without human intervention, achieving a perfect score in ExploitBench. To mitigate risks, OpenAI introduced a monitoring mechanism called the alignment monitor, which watches over the model's behavior and reasoning to detect and prevent potential malicious actions.

Initial access to the most sensitive cybersecurity functions will be limited to a program called Daybreak Blue, consisting of specialized cybersecurity organizations. Anthropic has also restricted access to its most advanced model, Claude Mythos, to a select few large enterprises specializing in cybersecurity to help identify and patch critical software vulnerabilities before malicious actors can exploit them.