OpenAI launches GPT-6 Astra under restricted high-risk controls

reports say that OpenAI launched GPT-6 Astra to selected enterprise customers. OpenAI describes it as a major step in reasoning, coding, vision and long-horizon agent work; its advanced cyber access is restricted because the company classifies it at the Critical cyber-capability threshold.

Why it matters: The launch combines broader autonomous capability with an unusually consequential security profile.

Business/technical/strategic impact: Enterprises should use isolated execution, deny-by-default network access, short-lived credentials, human approval for consequential actions and independent acceptance testing. Benchmark, alignment and performance claims remain vendor-reported.