Full Summary
This Wednesday morning, the U.S. government is moving to test advanced AI models from giants like Google, Microsoft, and xAI *before* they are released to the public. Both Reuters and The Washington Post confirm this initiative, led by the Commerce Department's National Institute of Standards and Technology, or NIST, aims to assess potential national security risks and societal harms. The push for pre-release vetting stems from President Biden's executive order on AI safety, with CNBC TV18 and SiliconANGLE reporting that Google, Microsoft, and xAI are joining OpenAI and Anthropic in voluntarily allowing these government checks. This shift replaces a previous "hands-off" approach to AI regulation, as news.cgtn.com notes, and is partly driven by concerns over Anthropic's new "Mythos" AI model. The News International reports Mythos is so powerful it can uncover tens of thousands of software vulnerabilities, including nearly 300 bugs in Firefox alone, leading Anthropic to deem it too dangerous for general release. The National Security Agency is already testing Mythos. What nobody expected: the European Union is also "bent out of shape" over the Mythos model, according to Let's Data Science, seeing it as a major policy stress test that could lead to increased EU regulation. Meanwhile, innovation continues, with DeepSeek V4 launching a 1-million-token AI model designed for affordability and transparency, and the University of Cape Town developing MzansiLM, an AI model that understands all 11 official South African languages, bridging a significant digital divide. Inworld AI's Realtime TTS-2 can now adapt its voice to your tone and emotions, making AI conversations feel more natural. This means your next interaction with advanced AI could be shaped by government oversight, while new AI tools are rapidly changing how we communicate and process information globally.