Anthropic DOESN’T Release A Model

On April 8, 2026, Tech Brew Ride Home spotlighted a pivotal moment in AI development: Anthropic's decision to withhold the release of its powerful new model, Mythos, due to profound security concerns. Instead of launching it publicly, Anthropic introduced Project Glasswing, a cybersecurity initiative that provides early access to the Mythos Preview model to over 40 critical infrastructure organizations—including AWS, Google, Microsoft, and NVIDIA—along with $100 million in usage credits and $4 million in donations to open-source security groups. The model demonstrated unprecedented capabilities, autonomously discovering and exploiting zero-day vulnerabilities in long-standing systems like OpenBSD, FFmpeg, and the Linux kernel, including chaining exploits to achieve full system control. Anthropic emphasized that the model’s power could destabilize global cybersecurity if released prematurely, and they’ve built a rigorous triage pipeline to responsibly disclose findings. Meanwhile, the rise of China’s open-source GLM 5.1 model—754 billion parameters, MIT-licensed, and capable of 1,700 autonomous steps—raises urgent questions about whether the global race to secure AI safety can outpace proliferation. The episode also covered Elon Musk’s amended lawsuit against OpenAI, seeking over $150 billion in damages to be directed to charity and demanding Sam Altman’s removal from the board, while the host shared a frustrating family trip to Las Vegas despite a rewarding visit to the Grand Canyon and Sedona.