Anthropic announced the limited release of Claude Mythos Preview, an AI model that can autonomously discover software flaws and generate working exploits. The company has placed the model in the hands of a select group of tech giants—including Microsoft, Apple, Google, and the Linux Foundation—through a consortium called Project Glasswing. Security experts say the system could dramatically lower the skill bar for creating multi‑stage exploit chains, prompting a reassessment of how organizations develop, patch, and defend software. Government officials are already discussing the potential fallout, underscoring the model’s far‑reaching implications. Read more →