My AccountNewsletters
National Post UK
OpenAI Retreats on Astra Project Amid Mounting Safety Fears

The San Francisco-based firm has abruptly cancelled its GPT-6.1 Astra model following rigorous internal testing. The dec

The National Post

HomeTechnology
Technology

OpenAI Retreats on Astra Project Amid Mounting Safety Fears

The San Francisco-based firm has abruptly cancelled its GPT-6.1 Astra model following rigorous internal testing. The decision signals a profound, if belated, shift in the industry’s race toward artificial general intelligence.

JM

Jack Morrison

Technology Editor

Published 11 October 2026·about 21 hours ago 3 min read
Text
Share

A Watershed Moment for Silicon Valley In a move that has sent tremors through the global technology sector, OpenAI confirmed on 30 September 2026 that it has formally abandoned the development of its much-anticipated GPT-6.1 Astra model. The decision follows a series of internal safety evaluations that reportedly identified catastrophic risks, forcing the organisation to pump the brakes on what was once the most ambitious project in its history. This sudden reversal serves as a stark reminder that even the most well-funded developers are struggling to reconcile the velocity of innovation with the necessity of existential safety.

The Internal Audit that Changed Course Sources familiar with the internal deliberations suggest that the Astra model exhibited unpredictable behaviour during red-teaming exercises, particularly regarding its capacity for autonomous reasoning in unconstrained environments. While OpenAI has kept the specific technical findings under lock and key, the decision to scrap the project entirely suggests that the flaws were fundamental rather than merely cosmetic. For a company that has defined the modern AI gold rush, this retreat represents a rare moment of introspection.

UK Industry Perspectives The impact of this decision is being felt keenly in Whitehall and across the UK’s burgeoning AI sector. The British government has recently prioritised the establishment of global safety benchmarks, and the OpenAI announcement is being interpreted by many as a vindication of the cautious regulatory framework advocated by the AI Safety Institute. Dr Helena Vane, a senior fellow in computational ethics at Imperial College London, notes that this represents a maturity shift. The industry is moving away from the 'move fast and break things' mantra of the previous decade towards a more sustainable, risk-averse model that prioritises long-term societal integration over quarterly progress reports.

A New Era of Precaution For years, the discourse surrounding AI has been dominated by a frantic pursuit of scale. The shelving of Astra suggests that the horizon for artificial general intelligence may be further away than previously touted by optimists. By prioritising safety protocols over aggressive release schedules, OpenAI is attempting to stabilise a volatile market that has grown increasingly concerned about the potential for widespread misinformation and system instability. As the dust settles, competitors are likely to be forced into a similar posture of reflection. The era of unchecked experimentation may well be coming to a close, replaced by a rigorous commitment to testing that, while slower, ensures the safety of the systems we are rapidly integrating into our critical infrastructure.

TopicsOpenAIArtificial IntelligenceTechnologySafetySilicon ValleyUK Tech
Share

About the Author

JM

Jack Morrison

Technology Editor

Jack Morrison reports for National Post UK with a focus on rigorous analysis and authoritative journalism — a trusted voice across British public life.

Reader Discussion · 5

Subscribers can join the conversation.

SUBSCRIBE TO COMMENT →
J
Julian Moore11 Oct

I wonder what this means for the UK tech sector's competitiveness. If we are constantly hitting these roadblocks, we might fall behind other nations that have a more relaxed approach to development.

F
Fiona Campbell11 Oct

Does anyone else find it worrying that the decision was made internally rather than by independent regulators? We are letting private corporations decide what is safe for the general public.

D
David Thorne11 Oct

A sensible decision. The pace at which these large language models were being pushed out was becoming unsustainable for any proper ethical oversight.

S
Sarah Jenkins11 Oct

I work in software development and honestly, this feels like a PR move rather than a genuine change of heart. They will just rebrand the same tech and release it under a different name in six months.

A
Arthur Penhaligon11 Oct

It is about time someone put the brakes on this runaway train. We have seen time and again that these companies are prioritising speed over basic human safety.

More from Technology

Most Read in Technology

Morning Briefing

The day's top stories, before 7am

Join 40,000 readers who start their morning with National Post UK.