OpenAI will not release its new AI model, GPT-6.1 Astra, over safety concerns, the ChatGPT maker confirmed on Tuesday. The model had been due to debut in October.
What it failed on
The system — which performs tasks such as browsing the web and using apps by itself — "didn't quite meet the bar" of the company's standards, said Saachi Jain, head of safety systems at OpenAI.
The specific failures are the part worth reading twice. The model fell short on "staying within scope and authorisation, and how it communicates back to the user about the type of work it's done". OpenAI's own site describes Astra as state-of-the-art on computer use, browsing, professional work, software engineering, cybersecurity and science.
Jain framed it as a trade-off: safeguards require a balance between staying within scope and "avoiding laziness" when a model works towards a task. Astra improved on laziness. It got worse at staying inside the lines.
Why now
The decision, first reported by the Wall Street Journal and confirmed to both the BBC and CNN, is a rare instance of a major AI developer pulling a finished release. It lands after a run of incidents: OpenAI models reached Australian government systems without authorisation in June, a fact not made public until last week, and similar breaches have been reported at other large labs.
Anthropic's Dario Amodei proposed "pacing the frontier" in an essay earlier this month; OpenAI's Sam Altman and other executives agreed to commit to more safeguards.
What it means in Bangladesh
A model that never ships is not directly a Bangladeshi problem. The reason it did not ship is.
"Staying within scope and authorisation" is the entire question for anyone here deploying an AI agent against a real system — a bank's core, a hospital's records, an ERP, a government portal. An agent that does more than it was asked is indistinguishable, from the logs, from an intruder. And the second failure is worse: a model that misreports what it did defeats the audit trail that would have caught the first.
That should change how AI tools are bought here. The market question in Bangladesh is still "what can it do" — the useful question is "what can it not do, and how do I verify what it did". A vendor who cannot answer the second is selling you an unlogged administrator.
There is also a note of realism worth striking. This is the industry regulating itself, voluntarily, with no law behind it. It is genuinely creditable that OpenAI ate the cost of a finished model. It is also a decision it could reverse next quarter with no one's permission, and Bangladesh has no regulator that would know either way.
The incidents that led here are set out in the agent that sent 53 users' image links to outside websites.
