Technology & Science
White House Alleges Moonshot AI Stole Anthropic Tech, Ponders Sanctions and New Coordination Bill
On 23 July 2026 Washington publicly charged Beijing-based Moonshot AI with covertly “distilling” Anthropic’s Fable model to build the 2.8-trillion-parameter Kimi K3 and warned that export controls and sanctions could follow, while Congress rushed out a bill letting U.S. AI firms coordinate on security responses.
Focusing Facts
- Science adviser Michael Kratsios posted that Moonshot ran large-scale, hidden distillation of Anthropic’s Fable, using Nvidia GB300 servers—including hardware located in Thailand—despite 2022 export bans.
- Treasury Secretary Scott Bessent said on 23 July that companies engaged in such “industrial-scale” IP theft could be sanctioned or placed on the Commerce Department’s Entity List.
- The same day, Senators Adam Schiff and Jim Banks introduced the Collaboration on Adversarial Threats and Security Risks Act, granting AI firms a narrow antitrust exemption to share threat data and delay risky model releases.
Context
The episode echoes the 1987 Toshiba-Kongsberg scandal—when U.S. anger over Japanese CNC machines allegedly aiding Soviet submarines triggered strict export controls—underscoring how dual-use tech repeatedly becomes a geopolitical flashpoint. Today’s AI clash sits atop longer arcs: the post-Cold-War faith in globalized innovation is giving way to techno-nationalist blocks, and the line between civilian and military software has blurred as models like K3 can write code, analyze imagery, and guide drones as readily as they draft emails. The U.S. response also reveals a structural tension: defenders of proprietary “closed” frontier models seek legal shields, while a growing open-weight movement argues that transparency, not secrecy, best secures critical systems. Whether Moonshot actually pirated Fable may matter less, on a century horizon, than the precedent Washington sets—each sanction or coordination carve-out tightens a feedback loop that could fragment the global AI knowledge commons much as the atomic secrets act did to physics in the 1940s.
Perspectives
US national security and mainstream political media
US national security and mainstream political media — They frame Moonshot AI's Kimi K3 as the product of large-scale theft of proprietary U.S. models and a direct security threat that justifies sanctions and new legislation. Coverage amplifies official talking points that favour American tech companies and hawkish China policy while offering scant technical evidence.
Open-source and decentralization advocates
Open-source and decentralization advocates — They hail Kimi K3’s forthcoming open-weight release as proof that transparent, inspectable AI can outclass closed U.S. systems and strengthen global innovation. Their enthusiasm for openness may underplay unresolved IP questions and portray Chinese progress as inherently liberating to advance their anti-walled-garden agenda.
Technical trade press and benchmark analysts
Technical trade press and benchmark analysts — They focus on whether Kimi K3’s headline numbers actually translate into superior performance, stressing that independent tests are needed and noting timeline inconsistencies in the U.S. accusation. By zeroing in on benchmarking nuance, they risk minimizing the geopolitical context and possible legal violations to maintain a posture of technical objectivity.
Like what you're reading?