LLMs are Getting a Lot Better and Faster at Finding and Exploiting Zero-Days

Opus 4.6, a large language model, is reported to be markedly better at discovering high-severity zero-day vulnerabilities in well-tested codebases out of the box—without fuzzers or custom tooling—indicating that AI can materially accelerate vulnerability discovery; the article is a news/analysis piece and does not disclose specific vulnerabilities, indicators, or confirmed active exploitation.