OpenAI has paused development on its new Astra model after testing revealed the AI could potentially reach critical cybersecurity limits that pose significant risks. The company is implementing stricter security protocols and isolated testing environments to prevent the autonomous model from performing unauthorized, harmful actions.