Radar Trends to Watch: May 2026 (www.oreilly.com)

In May 2026, the AI/ML landscape saw significant shifts as two major players diverged in their approaches to AI security. Anthropic limited access to its Claude Mythos model through Project Glasswing, focusing on a controlled rollout, while OpenAI made its GPT-5.5 widely available, raising concerns about security risks with its "open access." The implications of these developments are profound as the time gap between identifying and exploiting AI vulnerabilities has virtually disappeared, prompting an urgent need for robust risk management strategies within AI development. Additionally, this period marked a noticeable transition in AI's operational capabilities, moving beyond simple language tasks toward automation and collaboration across enterprises. The rise of open-weight models is reshaping the economic landscape in AI, with companies like DeepSeek and Moonshot Labs offering increasingly capable alternatives at lower costs. Notably, Google's release of 1,302 generative AI use cases and Anthropic's advancements in multimodal models further illustrate this evolving ecosystem. As developers navigate the growing array of technologies and tools, the focus will shift to understanding trade-offs between cost, support, and capability, highlighting the critical role of interoperability in future AI applications.