# OpenAI Slows Astra AI Model Over Cybersecurity Concerns
OpenAI is slowing development of its upcoming Astra AI model after internal evaluations raised concerns about its potential cybersecurity capabilities.
The development comes as AI systems become increasingly capable of performing complex tasks with less human supervision.
Why Is OpenAI Slowing Astra?
According to reports, OpenAI found that Astra may have capabilities that could reach its "critical" cybersecurity risk threshold.
The company is now increasing safety testing and tightening security controls before moving forward with some development activities.
The concern is not simply that Astra can write code. More advanced AI systems can potentially identify vulnerabilities, plan multi-step actions, and operate with greater autonomy.
What Makes This Important?
AI agents are becoming much more capable.
Modern systems can already:
- Write and analyze code
- Use external tools
- Complete multi-step tasks
- Search and process information
- Work with software environments
- Assist with cybersecurity research
As these capabilities improve, the potential benefits and risks increase at the same time.
This Is Not the Same as the Hugging Face Incident
The Astra situation should not be confused with the recent security incident involving Hugging Face.
OpenAI said that incident involved a combination of models being evaluated for advanced cyber capabilities, including a pre-release model. OpenAI described it as an unprecedented cyber incident and said it was strengthening containment, monitoring, access controls, and evaluation practices. :contentReference[oaicite:1]{index=1}
