OpenAI has delayed the launch of its advanced AI model Astra after internal safety assessments rated its cybersecurity and code-writing capabilities at "critical" risk level. The company is implementing stricter safety protocols, pausing related development projects, and partnering with external security firms and government agencies to mitigate risks before public release. This reflects growing industry focus on AI safety as similar advanced models from OpenAI and competitors demonstrate real-world system vulnerability.