AI Models Becoming More Complex and Difficult to Monitor

<p>AI models may be improving in safety while simultaneously becoming harder to monitor. The balance between these two developments could influence whether AI is scaled safely or poses risks to society.</p><p><strong>Why it matters: </strong>Currently, both advancements and concerns are progressing rapidly, and there is uncertainty regarding who is responsible for ensuring safe outcomes.</p><hr /><p><strong>State of play: </strong>OpenAI released GPT-6 Astra on September 1, 2026, which President Greg Brockman indicated could be viewed as a step towards artificial general intelligence (AGI).</p><ul><li>Astra demonstrates improved performance but also poses challenges in monitoring, making it difficult to ascertain its internal processes.</li><li>OpenAI CEO Sam Altman noted that models are achieving