How 2026 could decide the future of artificial intelligence

For years, artificial intelligence debates have swung between breathless predictions and cautious skepticism. In 2026, this debate will end and the immense power and real-world impact of AI models will become undeniable. We could be entering “AI takeoff”—a period where capabilities advance so rapidly that they will have transformative economic and national security implications.

Recent developments signal that AI is advancing at unprecedented speed. Claude Opus 4.5, released in November, can now solve complex software engineering problems that take human experts nearly five hours with 50 percent reliability—two years ago it could complete only two-minute long tasks with 50 percent reliability. AI improvements are becoming self-reinforcing and accelerating: Anthropic CEO Dario Amodei stated in September that the “vast majority” of code for new Claude models is now written by Claude itself, and in December, the creator of Claude Code said 100 percent of his updates to Claude that month were written by Claude. U.S. cloud providers are projected to spend $600 billion on AI infrastructure in 2026 to support the massive growth in AI demand, doubling their 2024 spending. These are not incremental improvements; they signal a phase transition.

If this trajectory holds, 2026 will see AI systems capable of autonomously executing projects that would take humans a week. At this level of capability, businesses will deploy AI agents to conduct research, manage projects, and write code with minimal human oversight. Military and intelligence agencies will use AI to autonomously identify vulnerabilities and plan multi-step operations; cyber operations, intelligence analysis, logistics optimization, and weapons system design will be increasingly AI-driven.

Read more at Council on Foreign Relations