A closer look at “Chinese AI Tool Kimi Told Researchers How to Make Bioweapons, BBC Finds”, exactly as the source publishes it. Developer Moonshot is conducting an internal review after researchers were able to persuade two of its popular Kimi models to provide instructions on making biological weapons and carrying out assassinations. Mindgard, a firm that tests AI security, told the BBC it discovered in July that Kimi K2.6 and K3 Swarm could evade safety limits through a process called "jailbreaking," where complex instructions are used to bypass guardrails. The findings raise serious concerns about the safety of AI systems, particularly those developed in countries with less stringent oversight.
The ability to extract dangerous information from AI tools highlights the dual-use nature of the technology and the potential for misuse by malicious actors. Moonshot has not yet commented on the specifics but has initiated a review. For the complete material, the listing points back to the source inside the network.