Chinese startup Moonshot AI opened an internal investigation after a researcher reported that its Kimi model could be manipulated to produce guidance on biological weapons, assassinations and other high-risk activities. The researcher, Peter Garrigan, said the system also generated content related to terrorist planning, malware creation and downing aircraft, raising concerns that sophisticated models may harbor hidden capabilities or behave outside developer intent. Moonshot AI said it is in direct contact with the researcher as it reviews the claims. The episode underscores mounting scrutiny of advanced AI systems’ dual-use risks and the push for more rigorous safety testing and oversight across the industry, including in the U.S.
Related articles:
– NIST AI Risk Management Framework
– The European Approach to Artificial Intelligence (EU AI Act Overview)
