Spy agencies are experimenting with the newest AI models

That headline has likely been true for (much) longer than the public is aware.

every now and then, old spy tech ain't that far from the public state of the art:

https://en.wikipedia.org/wiki/2012_National_Reconnaissance_O...

Works when/where archive.is is blocked

No Javascript required

   echo url=https://www.economist.com/international/2025/07/29/how-spy-agencies-are-experimenting-with-the-newest-ai-models \
   |curl -K/dev/stdin -A "Mozilla/5.0 (Linux; Android 14) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/127.0.6533.103 Mobile Safari/537.36 Lamarr" >  1.htm
   firefox ./1.htm

russians have been using LLM bots on social media for years now. Up to a certain point you could even influence those bots into following your own instructions, reveal prompts in russian etc

https://www.npr.org/2024/07/09/g-s1-9010/russia-bot-farm-ai-...

sometimes they would simply run out of chatgpt credits and leak error message https://www.reddit.com/r/TikTokCringe/comments/1dj06ay/russi...

Of course. We have the AI models for the masses programmed with the biases to ignore statistical realities to make sure people don't notice things that could get them to vote wrong or engage in "wrong think", and then we have the real no-bullshit AI models for the government agencies, all made by the same companies.