Notable organizations disclosed ~2,500 high- and critical-severity CVEs in July 2026, about 5× the pre-Mythos monthly record. Epoch AI's updated breakdown of vulnerability disclosures following Anthropic's Project Glasswing.

Even with automated AI R&D producing millions of virtual researchers, progress may be capped by our ability to divide, coordinate, and recombine their work. Epoch AI on why parallelization technology is a missing parameter in intelligence explosion models.

Expert assessments and cyber benchmarks led us to expect that frontier models were capable of executing this kind of cyberattack

We tested Pangram, GPTZero, and Originality.ai on human and AI text. All three caught nearly every passage written from a simple prompt, but missed roughly one in five passages imitating a specific author's style.

AI compute draws tens of gigawatts globally—comparable to New York state—yet a single chatbot query uses less than a microwave in 10 seconds. Epoch AI's guide to AI energy use, from training to inference to local grid impacts.

A look at the specialized hardware driving modern AI — why chips cost tens of thousands of dollars each, and why demand continues to outstrip supply.

The share of working days on which OpenAI's Codex collaborators merged very-high-effort code rose from about 2% in Q2 2025 to about 8% in Q2 2026.

Why we should think a little harder about what it takes to build a Dyson Sphere

Notable organizations disclosed ~1,300 high- and critical-severity CVEs in June 2026, roughly 3.5× the pre-Mythos monthly record. Epoch AI's breakdown of vulnerability disclosures following Anthropic's Project Glasswing.

GPT-4 topped Epoch's Capabilities Index (ECI) for roughly a year after its March 2023 release — no model since has led for as long. The second-longest reign, by o1, lasted barely three months.