OpenAI’s ambitious Project Astra has hit a major development wall.

Security teams flagged critical vulnerabilities, including unauthorized access and prompt injection risks, forcing a strategic slowdown.

While the AI promised seamless multimodal execution, OpenAI refuses to ignore potential "critical cyber capabilities." This safety-first pivot now threatens the highly anticipated enterprise rollout, proving that even futuristic agents must survive strict security gates.