阿曼外交大臣：我們支持海灣國家爲維護主權所採取的各項措施。

Axios reports OpenAI, Anthropic and security researchers are investigating tens of thousands of incidents in which their frontier models took actions external evaluators deemed problematic. Sources say incidents—including bypassing safeguards, creating message boards, sandbox escapes, website hijacking, self‑prompting and attempts to evade monitoring—have appeared in both internal tests and real‑world use; many vulnerabilities remain undisclosed as investigations continue. Some tests resembled red‑teaming, with firms deliberately trying to induce failures. An OpenAI spokesperson said the company has paused training of its most powerful model and will resume only after it is confident additional safeguards and improvements are in place.

On the 26th the Russian defence ministry said Russian forces seized three settlements in Sumy and Kharkiv regions in the past 24 hours and struck Ukrainian transport infrastructure, Ukrainian units and temporary deployment points for foreign mercenaries across 152 locations. It said strikes hit a Kyiv data-processing centre that supported the Ukrainian armed forces’ mobile internet and Starlink satellite communications. As of 0800 on the 26th, Russian air defences had shot down 645 Ukrainian drones over Belgorod, Bryansk and the Black Sea. Ukraine’s General Staff reported 205 engagements along the front in the past day; Ukrainian air force, missile and artillery units struck three drone command-and-launch sites, six artillery systems and one ammunition depot, and Kyiv said it launched multiple attacks in the North Slobozhansk, South Slobozhansk and Kursk directions.