OpenAI has officially reached a major milestone with the public release of GPT-6 Astra. Described as a 'generational leap,' Astra is the first model to achieve a 'Critical' cybersecurity capability rating under OpenAI's Preparedness Framework. This designation signifies the model can autonomously discover and exploit zero-day vulnerabilities, leading OpenAI to implement unprecedented 'recurrent depth' architecture safeguards that have sparked a fresh industry debate over reasoning transparency and safety monitoring.
As enterprises shift from experimental pilots to production-grade 'agentic AI,' the industry is grappling with a new economic reality known as the 'Inference Tax.' New reports indicate that 83% of organizations are finding their legacy infrastructure unable to sustain the massive memory and context requirements of autonomous agents, which often trigger hundreds of downstream actions from a single prompt, leading to skyrocketing data egress and storage costs.
As of September 2026, cumulative global spending by Big Tech on AI infrastructure and data centers has officially surpassed the $1 trillion mark.
Which AI model became the first to be designated with a 'Critical' cybersecurity capability tier in September 2026?