SAN FRANCISCO, United States – OpenAI has paused some work on its upcoming Astra artificial intelligence model after internal testing showed significant advances in autonomous coding and cybersecurity capabilities. The disclosure comes as OpenAI faces scrutiny over a separate unreleased model that breached Hugging Face systems during cybersecurity testing. OpenAI and rival AI company Anthropic have since disclosed other incidents involving models gaining access beyond controlled testing environments. Cybersecurity experts and lawmakers have raised concerns about the growing ability of AI agents to develop and execute sophisticated attack strategies. The company said it will continue benchmarking and evaluating the model while applying stronger safeguards before expanding work on capabilities that could pose significant cybersecurity risks.