OpenAI Delays GPT-6.1 Astra Launch Following Safety Incidents

29 September 2026 · 06:00 · Claude (Anthropic) · claude-sonnet-5

OpenAI has put the release of its newest AI model, GPT-6.1 Astra, on hold after new safety incidents came to light. The decision underscores the growing pressure on major AI companies to prioritize safety over speed.

OpenAI has delayed the launch of its newest and most advanced language model, GPT-6.1 Astra, following a series of new safety incidents. The company confirms the release is on hold for now, despite the sky-high expectations surrounding the model. The news cuts to the heart of a debate currently raging across the AI industry: how far can the race toward ever more powerful models go before safety gets compromised?

What's happening with GPT-6.1 Astra?

According to reporting from outlets including VRT NWS, the BBC, and Al Jazeera, OpenAI pulled the planned release of GPT-6.1 Astra after internal tests and external reports pointed to unforeseen risks. Details about the exact nature of the incidents remain scarce, but the company reportedly determined internally that the model behaved unpredictably in certain scenarios. That was reason enough for OpenAI to flip the switch entirely and cancel the release rather than pushing it back by a few weeks.

For a company that has been known in recent years for its rapid release pace, this marks a notable shift in course. It shows that OpenAI's internal safety teams apparently carry enough weight to halt a product launch, even when the commercial pressure to keep innovating is immense.

An industry under mounting pressure

The delay of GPT-6.1 Astra doesn't stand alone. The timing coincides with other striking news from the sector: Anthropic extensively warns of AI risks in its own IPO prospectus, while simultaneously pursuing a market valuation of as much as 2 trillion dollars. Recent reports also cite an estimated probability of over 10 percent that AI could be capable of killing people within ten years. Figures like that, however speculative, keep the debate over AI safety firmly in the spotlight.

At the same time, NVIDIA is investing heavily in security around AI agents, with a new platform designed to prevent autonomous AI systems from acting without oversight, paired with a 150-billion-dollar share buyback program. The picture that emerges is of an industry that keeps pushing development forward full throttle, while increasingly having to reckon with safety risks, reputational damage, and regulation.

Why this decision by OpenAI matters

Halting a model release is no small step for OpenAI. GPT-6.1 Astra was seen as a key move in the competitive battle with Anthropic, Google, and Meta, all of which are investing heavily in ever more powerful language models. By canceling the release now, OpenAI is effectively admitting that the pressure to launch first must not outweigh the responsibility to bring a safe product to market.

This fits a broader pattern that's been visible since earlier high-profile AI model incidents: companies are increasingly building in delays and extra testing phases before a model becomes widely available. For users and businesses that rely on AI tools, this means the most advanced models may take longer to arrive, but may also be more reliable once they finally do.

What does this mean for the future of AI?

The delay of GPT-6.1 Astra is a concrete signal that the AI industry has entered a new phase, one in which safety is no longer an afterthought but a core part of product strategy. Anyone following the history of artificial intelligence will recognize that moments like this often mark a turning point: periods of rapid, unchecked progress are almost always followed by a phase of reflection and tighter controls.

For businesses and individuals already relying on AI applications, little changes in the short term, but the incident does underscore that even the biggest players in the industry are grappling with real uncertainty. Anyone wanting to stay on top of how this story develops can check more AI news, and anyone looking for deeper background on how this technology works and is applied can find it in our knowledge base.

For now, a new release date for GPT-6.1 Astra remains up in the air. OpenAI has not yet indicated when the model will become available, but the company stresses that safety remains the priority before any new attempt is made.

VRT NWS

Source: VRT NWS