Moonshot AI curbs cyberattack capabilities in Kimi model launch

Traders Union
All News
Business & Wealth
Companies
Moonshot AI curbs cyberattack abilities

Highlights

Moonshot released Kimi K3.1 Pro, a 3-trillion-parameter-class open-weight model, with cyberattack capabilities curtailed by filtering cyber-related training data.
Kimi K3.1 Pro scores 81.0% on internal cyber benchmarks versus 90.5% for an unrestricted version, and offers a 1-million-token context window plus a faster Highspeed variant.
China's draft export controls target AI-driven malware-generation and vulnerability tools, while Moonshot plans to release Kimi K3.1 Pro's open weights by November 4.

Model safeguards and release timeline

As reported by Reuters, Beijing-based Moonshot announced the restrictions as it introduced Kimi K3.1 Pro, which it describes as its most capable 3-trillion-parameter-class open-weight model. The company says it plans to publish the model's weights by November 4, allowing outside developers to run and modify it on their own systems.

Moonshot says it removed cyber-related data from the training corpus and reinforcement-learning environments used to develop the model, while keeping general coding skills. The company says it does not aim to release open weights that would extend the absolute frontier of cyber capability.

K3.1 Pro scores 81.0% on Moonshot's internal cyber benchmark, compared with 90.5% for a similar version without the risk-reduction measures, according to the company. The test involves finding and exploiting software vulnerabilities in a controlled environment, and Moonshot says the comparison shows filtering cyber-related data reduces offensive cyber performance, although the model still outperforms its predecessor; the results have not been independently verified.

K3.1 Pro has a 1-million-token context window, allowing it to process very large amounts of text at once. Moonshot also releases a faster version, K3.1 Pro Highspeed, which it says runs at two to three times the speed of the standard model.

Cybersecurity concerns shape AI oversight

Governments, technology companies and security researchers are growing increasingly concerned that more autonomous AI could make sophisticated hacking easier and cheaper. Models that can write code, operate software tools and complete long, multi-step tasks may help users identify weaknesses in programs and automate parts of cyberattacks.

China proposed export controls last week that would ban exports of AI-driven malware-generation and mutation technology and restrict AI-assisted vulnerability-discovery tools. The draft measures, which remain open for public comment, cite the risk that such capabilities could be used against critical infrastructure, defence-industrial and core government systems, or to bypass security protections.

Moonshot says open-weight models can also improve security by helping more organisations inspect and secure their own software, including older systems that receive limited attention. But open-weight releases make central oversight harder than with proprietary, closed-weight systems because once the files are widely distributed, the original developer has less ability to restrict access, monitor use or update safeguards.

AI-driven cyber threats are pushing banks and regulators to strengthen operational resilience, our earlier coverage noted. In particular, the Bank of England urged major lenders to plan for “bare metal” recovery options that can keep critical services running if primary systems are compromised, warning that advances in AI increase the scale and complexity of attacks.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.