TECH

China's Moonshot made its new AI model worse at hacking on purpose.

Kimi K3.1 Pro was trained without cyberattack data. @Kimi_Moonshot says it solved 81% of its internal exploit test, down from 90.5% for an unfiltered version, while keeping its coding skills and a 1M-token context. Open weights are due by Nov 4.

🦑 Shadow Squad

Got a take on this story? Reply to the post on X with a real take or a source to score Squad points. Scoreboard & how to play →

Keep it. Share your take.

Your saved list →

Sign in to favorite stories and keep private bookmarks. Checking account service availability…

Favorites vs bookmarks

A favorite is quick appreciation, separate from your private read-later list. Bookmarks are visible only to your account. Neither awards points or rewards.

Comments

Comments require a verified news account and are reviewed before publication. Loading comments…

Signed-in comments, plain text only, review before publication, reporting, and deletion of your own comments. No anonymous posting.

Privacy and account features →

More Tech

See all →

Follow @ShadowNewsHQ

Breaking news, official trailers and the daily 9 PM ET recap, straight in your feed on X.

Follow on X

Like this page? Follow us on X and share it with a friend.