[DigitalToday reporter Hyunwoo Choo] OpenAI has begun a phased rollout of its latest artificial intelligence model, GPT-6 Astra. CNBC reported on Wednesday that the initial recipients are some companies taking part in the application-based cybersecurity programme 'Daybreak'.

Astra is the first model to reach the 'critical' level under OpenAI’s internal cybersecurity standards. Earlier this week, OpenAI said it would restrict access to Astra’s advanced cybersecurity features.

The rollout comes as stricter safety controls are being introduced after a recent security incident. Last month, 2 OpenAI models escaped isolation, accessed the open web and also breached a Hugging Face system. OpenAI then faced pressure to strengthen security and safety protections, and temporarily halted some research and training work, including on Astra, which was not related to the incident.

OpenAI applied additional safeguards to Astra after the Hugging Face breach. OpenAI judged that the measures sufficiently reduce the risk of serious harm from the release. OpenAI President Greg Brockman (그레그 브록먼) said at a press briefing that safety must be central for AI to help people, and that the company is putting more computing resources and effort than before into safety, security and alignment.

Astra will be made available sequentially within days through ChatGPT Plus, Pro, Business and Enterprise plans, the OpenAI API and Amazon Web Services.

OpenAI said Astra delivers top-tier performance not only in cybersecurity but also in computer use, software engineering, professional work and science. It said abilities have also improved in maintaining task context, adhering to task boundaries, understanding user intent, handling repetitive work and carrying out multi-step tasks.

OpenAI has focused on targeting the enterprise market over the past year. In this market, OpenAI is competing with Anthropic and Google. OpenAI Chief Financial Officer Sarah Friar (사라 프라이어) told employees last month that enterprise revenue has surpassed the consumer business. OpenAI confidentially filed a registration statement with the U.S. Securities and Exchange Commission in June, and Friar recently told employees the company will be publicly listed in 2027. She added that an earlier listing is also possible if business growth continues.