OpenAI has delayed the wider release of GPT-5.6 following a U.S. government request over national security concerns. Here's what happened and why it matters for the future of AI.

In early 2026, Anthropic explored an unconventional direction in AI evaluation. Instead of focusing only on intelligence, their advanced model preview Claude Mythos was placed into a structured psychiatric style interaction. This was not a standard test. It was designed to examine whether an AI can maintain psychologically consistent behavior when engaged over time in a human like format. A licensed psychiatrist conducted around 20 hours of structured sessions across multi