Six Chinese AI firms accused of aggressively copying US frontier models

Key Points:

• The US government has accused six Chinese AI firms, including DeepSeek, Moonshot AI, and Alibaba, of conducting industrial-scale attacks to extract proprietary capabilities from leading US AI models like Claude, GPT, Gemini, and Grok since late 2024, allegedly with Chinese government awareness.
• These firms reportedly use methods such as exploiting AI model APIs with fake accounts and prompt injection techniques to "jailbreak" models and obtain internal reasoning, enabling China to reduce AI development costs and timelines significantly.
• US agencies recommend AI companies enhance detection of suspicious activities, implement user identity verification, and subtly degrade model responses for suspected malicious users to deter theft, though these measures may frustrate legitimate users due to reduced model performance or altered outputs.
• The US emphasizes the need for coordinated international cooperation and information sharing among AI firms and allied governments to effectively combat ongoing intellectual property theft and protect the US lead in AI development.
• China denies the allegations, calling them groundless and politically motivated, while affirming its AI advancements stem from scientific self-reliance; Chinese officials urge the US to focus on cooperation rather than accusations amid rising tensions ahead of a scheduled Trump-Xi meeting.