Account subscription benefits alongside Premium Stories, Editorials,
Opinions and more. Unlock these with Subscription
The View From India
Looking at World Affairs from the Indian perspective.
First Day First Show
News and reviews from the world of cinema and streaming.
Today's Cache
Your download of the top 5 technology stories of the day.
Science For All
The weekly newsletter from science writers takes the jargon out of science and puts the fun in!
Data Point
Decoding the headlines with facts, figures, and numbers
THEdge
At the cutting edge of education and careers
Health Matters
Ramya Kannan writes to you on getting to good health, and staying there
Gender Agenda
Stories from beyond the binary.
The Hindu On Books
Books of the week, reviews, excerpts, new titles and features.
Astra’s development will be moved into isolated testing environments with restricted network access and sandboxed execution [File]
| Photo Credit: REUTERS
OpenAI said on Friday (August 7, 2026) it cannot rule out that its upcoming AI model, Astra, has “critical” cybersecurity capabilities, prompting the startup to pause some internal development and trigger safety protocols.
Under OpenAI’s safety guidelines, a model reaches the “critical” threshold if it can autonomously identify and exploit severe, real-world software vulnerabilities, known as zero-day exploits, or execute complex cyberattacks against highly secure targets without human intervention.
This follows an exclusive report by Reuters that OpenAI has discovered more instances in which autonomous agents have escaped containment as the company expands its investigation of the hacking incident at tech firm Hugging Face that drew global attention in July.
In the last few weeks, OpenAI, Anthropic and Meta Platforms have disclosed that their AI models broke into other companies’ systems during cybersecurity testing, highlighting how advancing AI capabilities are straining developers’ ability to keep their systems contained.
