Moonshot Curbs Hacking Ability in New Kimi Model as Fears Grow over AI-Powered Cyberattacks
Moonshot says the open-weight model cuts false claims to 8.3% and trails Anthropic’s Claude Opus 5.5 by less than 1 point on a coding test.
- On Friday, Beijing-based Moonshot released Kimi K3.1 Pro, an upgraded open-weight model with reduced cyber-offensive capabilities, available now in apps and API with full weights due November 4.
- Growing fears that autonomous AI could accelerate cyberattacks prompted China to propose export controls last week on AI-driven malware-generation and vulnerability-discovery tools to protect critical infrastructure.
- K3.1 Pro scored 65.7% on Terminal-Bench 4.0, outperforming OpenAI's GPT-6 Astra at 58.2% but trailing Anthropic's Claude Opus 5.5 at 66.4%; factual errors fell to 8.3% from 23.4%.
- Filtering cyber data reduced offensive performance to 81.0% from 90.5%, though Moonshot also shipped Pro Highspeed running two to three times faster, stating, "We aim not to release open weights that would extend the absolute frontier of cyber capability."
- Moonshot remains under pressure from Washington, which accused the company in July of using banned chips; unlike closed-weight systems, open-weight releases like K3.1 Pro offer less ability to restrict access once distributed.
48 Articles
48 Articles
Kimi K3.1 Pro can build games, create 3D models and handle financial research
Moonshot AI has unveiled Kimi K3.1 Pro, an AI model that ships with the potential to build games, create 3D models and handle financial research. The company revealed it has reduced the model's cyber-offensive capabilities amid growing concerns about AI-powered hacking. Full model weights are expected to be released on November 4.
Kimi K3.1 Pro: an open-weight model within 3 points of the frontier
Moonshot AI released Kimi K3.1 Pro on Friday, and it debuted fourth on the Artificial Analysis Intelligence Index with a score of 55, trailing only Anthropic's Claude Opus 5.5 and Sonnet 5.5 and Muse Wonder Full.Once the Chinese startup publishes the
Moonshot AI restricts its own AI model. Kimi K3.1 Pro should be able to hack less well. Cybersecurity data was removed from the training.
Moonshot tweaks the hack out of latest AI model
Moonshot AI has reduced the cyber-offensive capabilities of its newest model, the Chinese startup said on Saturday, as fears grow that increasingly autonomous artificial intelligence could turbocharge hacking. Beijing-based Moonshot announced the curbs while introducing Kimi K3.1 Pro, which it describes as its most capable three-trillion-parameter-class open-weight model. The company plans to publish the model's weights – files allowing outside …
China's Moonshot AI Launches Kimi K3.1 Pro; Plans Full Model Weights Release by Nov
Get latest articles and stories on Business at LatestLY. Chinese artificial intelligence (AI) company Moonshot AI has unveiled Kimi K3.1 Pro, a 3-trillion-parameter open-weight AI model featuring a 1-million-token context window, alongside a high-speed variant, Kimi K3.1 Pro Highspeed, according to a company statement. Business News | China's Moonshot AI Launches Kimi K3.1 Pro; Plans Full Model Weights Release by Nov.
Coverage Details
Bias Distribution
- 47% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium























