Anthropic's 'Dangerous' AI, Meta's Muse Spark & OpenAI Shelves UK Project

This week on AI RADAR: **Anthropic** reveals **Claude Mythos Preview**, a model deemed too dangerous for public release due to extreme cybersecurity risks, launching **Project Glasswing** instead.We also cover **Meta's** new **Muse Spark** model from its Superintelligence Labs, and why **OpenAI** shelved its multi-billion dollar **Stargate UK** infrastructure project over energy constraints.Plus: **Google Gemini** integrates "Notebooks" for deep research and **MIT** develops a technique to shrink AI during training.Sources:• Anthropic: https://www.anthropic.com/glasswing• Meta: https://www.theguardian.com/technology/2026/apr/09/meta-debuts-new-ai-model-in-first-test-of-costly-superintelligence-team• OpenAI: https://www.theguardian.com/technology/2026/apr/09/openai-shelves-stargate-uk-in-blow-to-britains-ai-ambitions• Google: https://www.pcmag.com/news/google-geminis-new-notebooks-help-organize-chats-research-into-projects• MIT: https://news.mit.edu/2026/new-technique-makes-ai-models-leaner-and-faster-while-theyre-still-learning