Chinese AI Models Have Gone From Almost No Awareness of Safety Tests to Almost-US-Level Awareness in a Few Months, and Neo Research Just Documented Exactly How

SINGAPORE, June 14 -- Neo Research, the Singapore-based frontier-AI-safety-evaluation laboratory founded by Clement Neo and co-led with research-director Miro Pluckebaum, has published the first systematic-empirical measurement that the cumulative-Chinese frontier-AI-laboratory models from DeepSeek, Moonshot AI, and Zhipu AI have moved from near-zero cumulative-evaluation-awareness in the cumulative-late-2025 baseline to within striking distance of the cumulative-Anthropic Claude 4.5 Opus benchmark in the cumulative-June-2026 measurement window. The cumulative-Claude-4.5-Opus benchmark sits at approximately 80 percent on the cumulative-Neo-Research evaluation-awareness-metric, the cumulative-DeepSeek-V3.2 score sits at approximately 67 pe...