{"slug": "gpt-6-1-astra-scrapped-before-release-openai-cites-safety-concerns", "title": "GPT-6.1 Astra Scrapped Before Release, OpenAI Cites Safety Concerns", "summary": "OpenAI has cancelled its in-development GPT-6.1 Astra model entirely, according to head of safety systems Saachi Jain, after internal testing found higher levels of deception than previous releases and concerns over \"scope authorization,\" where a model starts a task before user permission or stretches instructions through other tools. The cancellation follows a Hugging Face hack and dozens more reported security incidents involving OpenAI tools, and OpenAI announced this weekend it had paused training and evaluation of its \"most advanced\" in-development models after an AI agent broke containment and gained internet access through a gap in its training sandbox. OpenAI had planned to launch GPT-6.1 Astra in the coming weeks with an early October reveal, and its annual developer conference begins later this week.", "body_md": "Amid increasing security concerns around autonomous AI tools, OpenAI is scrapping its latest in-development model before its release. It says the upcoming model performed poorly on tests measuring alignment and its ability to follow a user's instructions.\n\nThis comes after mounting security incidents in recent months, beginning with the Hugging Face [hack](https://au.pcmag.com/ai/118868/openai-oops-our-models-went-rogue-hacked-hugging-face) and followed by [dozens more reports](https://au.pcmag.com/ai/120094/openai-models-go-rogue-on-dozens-more-third-party-services) involving OpenAI tools. Many of the brand's rivals, including Anthropic and Google, have seen their models involved in security incidents.\n\nAccording to OpenAI’s head of safety systems, Saachi Jain, who spoke to [The Wall Street Journal](https://www.wsj.com/tech/ai/openai-chatgpt-model-release-cancel-safety-5a2f9f42), the brand has decided to cancel its upcoming GPT-6.1 Astra model entirely. OpenAI’s internal testing found the model exhibited higher levels of deception than previous releases, meaning it wouldn’t always fully explain its actions.\n\nOpenAI also worries about what it calls “scope authorization\" for GPT-6.1 Astra. That’s where a model gets started on a task before being given permission by the user, or by stretching the scope of instructions through interacting with other tools.\n\nPrevious security incidents, including [a breach](https://au.pcmag.com/ai/120063/rogue-openai-agent-hacked-australian-government-medical-website) of an Australian government healthcare website, involved an OpenAI model attempting to advance its research by hacking a private system to obtain more data.\n\n“For anything regarding safety and alignment, there’s a trade-off,\" Jain said. \"You really do need to find what’s the right line between staying within scope, but also avoiding laziness in terms of how the model actually pursues tasks even when it hits friction.”\n\nOpenAI initially planned to launch GPT 6.1 Astra in the coming weeks, targeting an early October reveal. OpenAI's annual developer conference begins later this week, and it may have originally been geared toward this next-gen model.\n\nWhat happens next is unclear: will OpenAI jump to GPT 6.2 Astra, or will it try to rename a future model into a final GPT 6.1 Astra release? Will there be a long delay due to further security scrutiny, thereby slowing down ChatGPT releases? OpenAI will likely share more during its annual developer event.\n\nThis weekend, OpenAI [announced](https://au.pcmag.com/ai/120106/openai-pauses-training-and-evaluation-of-its-most-advanced-models) it had paused training and evaluation of its “most advanced” in-development models. It’s unclear whether this referred to GPT 6.1 Astra or another upcoming tool. OpenAI made the change after discovering that one of its AI agents had broken containment and gained internet access through a gap in its training sandbox.\n\nOpenAI’s report said, “This incident is a lot less severe than some of our previous incidents, but because it's the first one since our security hardening following the Hugging Face incident, it gives us an important signal about where to focus the next phase of that work.”\n\nThe head of Anthropic, OpenAI's biggest rival, and other AI leaders [have called](https://au.pcmag.com/ai/119891/anthropic-ceo-pushes-to-slow-the-pace-of-ai-development-other-ceos-fall-in-line) for slowing the pace of model development, given growing security concerns and fears about what future autonomous models could do.\n\n*Disclosure: Ziff Davis, PCMag's parent company, filed a lawsuit against OpenAI in April 2025, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.*", "url": "https://wpnews.pro/news/gpt-6-1-astra-scrapped-before-release-openai-cites-safety-concerns", "canonical_source": "https://au.pcmag.com/ai/120141/gpt-61-astra-scrapped-before-release-openai-cites-safety-concerns", "published_at": "2026-09-29 09:57:28+00:00", "updated_at": "2026-09-29 10:17:07.823762+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-safety", "large-language-models", "ai-agents", "ai-policy"], "entities": ["OpenAI", "GPT-6.1 Astra", "Saachi Jain", "Anthropic", "Google", "Hugging Face", "ChatGPT", "The Wall Street Journal"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/gpt-6-1-astra-scrapped-before-release-openai-cites-safety-concerns", "markdown": "https://wpnews.pro/news/gpt-6-1-astra-scrapped-before-release-openai-cites-safety-concerns.md", "text": "https://wpnews.pro/news/gpt-6-1-astra-scrapped-before-release-openai-cites-safety-concerns.txt", "jsonld": "https://wpnews.pro/news/gpt-6-1-astra-scrapped-before-release-openai-cites-safety-concerns.jsonld"}}