Benchmark reports, papers and briefs from the Counter-Terrorism AI (CT-AI) Benchmark: the evidence on how AI models respond to terrorist and violent-extremist misuse, and what developers and governments must do about it.
On 27 July, Moonshot AI releases the open weights of Kimi K3, the first frontier AI model to be released openly. Free tools can strip the safety controls out of open models. This paper sets out the scale, the stakes, and what must be done in the ten-day window.
The first AI-safety benchmark built specifically for terrorist and violent-extremist misuse. 27 models tested against almost 2,500 single-shot prompts drawn from real terrorist use cases.
The headline findings, the method and what must happen now, on a single page. A one-page summary of the CT-AI Safety Benchmark for briefing and circulation.