A. DeepSeek
B. Moonshot AI
C. Baidu
D. Alibaba
Explanation:
Moonshot AI had removed cyber-related data from the training corpus and reinforcement-learning environments for Kimi K3.1 Pro. The company said it wanted to preserve general coding ability while reducing the model’s ability to assist with cyberattacks. Its reported internal benchmark showed a lower score on vulnerability discovery and exploitation after these changes.
Important points
- – Kimi is a family of AI models developed by Moonshot AI.
- – Open-weight models allow developers to download and run model weights.
- – Once model weights are widely distributed, it can be harder to control how a model is used.
Related questions and answers
– Q: What is a reinforcement-learning environment?
A: A setting in which a model learns through actions, feedback and rewards.
– Q: What is an open-weight AI model?
A: A model whose trained weights are made available for others to download or use, subject to its licence.
– Q: What is a software vulnerability?
A: A weakness that may be exploited to compromise a system.