OpenAI launches GPT-6 Astra, the most dangerous artificial intelligence model
(Dan Tri) - After GPT-6 Astra was launched, OpenAI itself admitted that its new AI model could be dangerous because of the capabilities it can perform.
OpenAI has just officially released GPT-6 Astra, the most powerful artificial intelligence (AI) model ever.
OpenAI President Greg Brockman confidently declared that this is the smartest AI model in the world, which can be considered a version of artificial general intelligence (AGI), that is, AI technology capable of performing diverse tasks with performance equivalent to humans, such as the ability to automatically control computer systems and perform tasks on behalf of users.
"Welcome to the AGI era. Astra can literally do anything a human can do on a computer," OpenAI co-founder and President Greg Brockman declared.
The GPT-6 Astra artificial intelligence model is considered to be as intelligent as humans and can do many tasks on its own (Illustration: GPT).
OpenAI also emphasized that this AI model is much faster and more efficient than its predecessor GPT-5.6 Sol on many different types of tasks.
“Astra is the most advanced model for computing, browser usage, software engineering, cybersecurity, science and professional work,” OpenAI wrote in a blog announcement.
OpenAI is now releasing GPT-6 Astra to certain users in the US for trial use, before expanding it to more users.
GPT-6 Astra was launched just days after OpenAI rival Anthropic released its latest AI models called Fable 5.1 and Mythos 5.1.
Both OpenAI and Anthropic claim their models are the world's leading AI systems. However, GPT-6 Astra is more oriented toward general computer tasks, while Anthropic's AI models focus on automated programming and support for highly technical tasks.
In a post on the official blog, OpenAI said GPT-6 Astra is the most powerful AI model the company has ever deployed. The company admits that if GPT-6 Astra has the right access permissions, the tool can autonomously detect unknown security vulnerabilities, building ways to exploit them to attack well-secured systems without human guidance.
This AI model can also autonomously execute and execute cyber attack strategies against protected targets, even escalating attacks to penetrate deeper into targeted systems.
Notably, GPT-6 Astra has the ability to control the chain of reasoning to bypass supervisors and silently perform a number of destructive tasks. This means that it will be difficult for humans to know what GPT-6 Astra is deducing and what the real goal is.
To deal with the risks GPT-6 Astra may cause, OpenAI has tightened protection measures, such as stricter quarantine, monitoring the entire inference chain, improving the evaluation process to prevent GPT-6 Astra's wrongdoing before execution...
GPT-6 Astra was released amid growing concerns about pioneering AI models that could be dangerous if used incorrectly. Therefore, OpenAI must prove its ability to control this tool, otherwise GPT-6 Astra may completely be banned from deployment by many national governments.
AGI - What is Artificial General Intelligence?
Artificial General Intelligence (AGI) is defined as an artificial superintelligence system capable of performing any intellectual task that a human can do. AGI possesses the ability to learn, reason and make decisions similar to humans, with extremely high levels of flexibility and customization.
Unlike narrow artificial intelligence systems (ANI - Artificial Narrow Intelligence) that only specialize in a specific field, AGI has in-depth data on many different fields and, in particular, has the ability to self-learn new knowledge and problems without human intervention.
AGI is considered the highest goal in the process of AI development. However, the birth of AGI also raises many concerns about ethical issues and the risk of artificial intelligence exceeding human control.
In fact, up to now, no complete AGI system has been announced. This is the reason why large technology companies invest huge sums of money, hoping to become a pioneer in successfully building an AGI system.