Moonshot limits the offensive capabilities of its Kimi K3.1 Pro model before releasing its weights
China's Moonshot AI announced it reduced the offensive capabilities of its Kimi K3.1 Pro model amid the risk that AI could automate hacking tasks. The company removed cybersecurity-related data from the training corpus and from reinforcement learning environments, while keeping general programming capabilities. The measure applies before distributing the model's open weights, whose release was scheduled before November 4, 2026, according to Reuters. Moonshot said it does not intend to distribute open-weight models that expand the frontier of offensive cybersecurity capabilities.
In 35 seconds
- Moonshot removed cybersecurity data from Kimi K3.1 Pro's training and reinforcement learning.
- The company kept general programming capabilities so as not to eliminate legitimate uses.
- The measure applies before the open weights are published, scheduled before November 4, 2026 according to Reuters.
Read the full story at Inspenet
IASummary generated by artificial intelligence from a story by Inspenet. It may contain errors: the original source prevails.
Comments
Be respectful. Comments reported by several people are hidden until a moderator reviews them.
You do not need an account to read, only to comment. Sign in with a code sent to your email or with your Google account. We only show the name you choose, never your email.