OpenAI rolls out powerful GPT-6 model with stronger safeguards amid rising AI cybersecurity fears

San Francisco, September 4 — The creators of ChatGPT, OpenAI, announced plans to roll out their newest and most advanced AI model, GPT-6, also known as Astra, to select customers. President Greg Brockman emphasized the importance of safety in the development of such powerful AI models. Since the launch of the previous GPT-5 model a year ago, concerns about the capabilities of advanced systems have grown, especially following some concerning incidents involving OpenAI and its rival, Anthropic.

After two OpenAI models were involved in a security breach at AI platform Hugging Face earlier this summer, the company decided to develop GPT-6 with enhanced safeguards, although Astra itself was not implicated in the hack. Some cybersecurity-focused customers will be able to access the new AI model starting Thursday, with a full rollout to other paying customers to follow.

However, users on free or low-cost plans will not have access. OpenAI aims to make GPT-6 available "as quickly as we can," despite user frustration. The company highlighted the model's ability to autonomously handle a wide range of demanding tasks, including website creation, scientific analysis, game development, cybersecurity, and coding.

Astra could reportedly save users up to 10 minutes when building autonomous AI agents, for example, reducing apartment hunting from six hours to under 10 minutes. OpenAI's Chief Scientist Jakub Pachocki acknowledged the uncertainty surrounding how a new model behaves once released, stating that while a model can become highly proficient at achieving a goal, it may still act against intended purposes.

OpenAI must be prepared to slow down or halt scaling when safety is insufficient. Despite these concerns, OpenAI CEO Sam Altman argued that the company is releasing the model at a critical moment, as the world is on the brink of a significant change in the cyber attack landscape. OpenAI claims that the only way society can collectively defend against new cyber threats is by using tools like Astra to rapidly respond.

In July, OpenAI confirmed that it has reached one billion active users across all its products. The move comes amid growing concerns about AI-related cybersecurity risks, with OpenAI, Anthropic, and over 100 other organizations signing a letter calling for coordinated global efforts to address these issues.