KARACHI – OpenAI plans to make its upcoming artificial intelligence model, Astra, generally available to the public, but concerns over its cybersecurity capabilities have delayed a wider release, CEO Sam Altman has said.
In a post on X, Altman described Astra as a powerful model and said OpenAI was working towards making it available more broadly.
He also argued that keeping powerful AI systems accessible only to a limited group was not a desirable approach.
Cybersecurity concerns delay wider release
The planned release comes as OpenAI assesses the potential risks associated with Astra’s capabilities. According to recent reporting by The Verge, OpenAI has paused some internal activities involving the model after evaluations showed significant advances in agentic coding and cybersecurity.
The evaluations reportedly raised concerns that Astra could potentially reach a level of cyber capability that OpenAI’s Preparedness Framework classifies as “critical”.
Read more: Chinese AI that shocked the world: Here’s what you need to know
Astra shows advances in scientific reasoning
Although Astra has not been officially launched, OpenAI has provided some indications of the capabilities of its internal version.
Earlier this month, OpenAI researchers said an internal version of Astra had made progress on 10 long-standing problems in mathematics, quantum complexity and theoretical computer science. The researchers presented the results as evidence of advances in scientific reasoning.
Reports on the research said Astra generated core arguments for the problems and helped formalise the results in Lean, a system that allows mathematical proofs to be checked mechanically.
Read more: Punjab Motorcycle Fines for Missing Mirrors and Indicators
| Key detail |
Information |
| Company |
OpenAI |
| AI model |
Astra |
| Planned availability |
General public release |
| Current status |
Public release delayed |
| Main concern |
Advanced cybersecurity capabilities |
| Reported capabilities |
Agentic coding and cybersecurity |
| Scientific progress |
Advances reported on 10 long-standing problems |
| Research areas |
Mathematics, quantum complexity and theoretical computer science |
| Proof formalisation |
Lean |
| Research collection |
249 pages |
| Public release date |
Not announced |
Researchers compiled the work into a 249-page research collection. However, these results come from an internal version of Astra and do not necessarily represent the capabilities of the model OpenAI may eventually release to the public.
No public release date announced
For now, OpenAI has not provided a specific date for Astra’s public launch. The company appears to be balancing its plans for wider access with additional safety evaluations, particularly in cybersecurity.
If Astra is eventually released as planned, its capabilities in coding, scientific reasoning and cybersecurity could make it one of OpenAI’s most significant model launches. For the moment, however, the timing of the release remains uncertain.
Read more: AI takes away entry-level tech jobs: Report
Key points:
- OpenAI plans to make its upcoming Astra AI model generally available to the public.
- CEO Sam Altman said the company does not believe powerful AI models should remain limited to a small group.
- OpenAI has delayed Astra’s wider release amid concerns over its advanced cybersecurity capabilities.
- Internal evaluations reportedly found major progress in agentic coding and cybersecurity.
- Astra has also shown promising results in mathematics, quantum complexity and theoretical computer science, although the model has not yet been publicly released.
