Trump Administration Targets Foreign Tech Exploitation in AI Sector, with China in Focus
The Trump administration has expressed its intent to intensify scrutiny over the exploitation of United States artificial intelligence models by foreign tech firms, prominently identifying China as a principal concern during a period when the Asian nation is increasingly competitive in the arena of AI.
In a memorandum released on Friday, Michael Kratsios, the chief science and technology advisor to the president, alleged that entities based “principally in China” are orchestrating deliberate, large-scale operations to extract capabilities from leading U.S. AI systems. He underscored concerns about their “exploitation of American expertise and innovation.”
According to Kratsios, the administration will collaborate with domestic AI companies to pinpoint such activities, fortify defenses, and explore punitive measures for violators.
This memorandum emerges amidst a growing challenge to U.S. supremacy in artificial intelligence, a sector the White House deems crucial for establishing global standards and seizing both economic and military advantages.
Michael Kratsios, White House Director of Science and Technology Policy, addresses the White House Task Force on Artificial Intelligence Education.
A recent report from Stanford University’s Institute for Human-Centred AI suggests that the disparity between U.S. and Chinese AI model performances has “effectively closed.”
In response, the Chinese embassy in Washington condemned the “unjustified suppression of Chinese companies by the U.S.”
Embassy spokesperson Liu Pengyu emphasized that China is committed to fostering scientific and technological progress through collaboration and fair competition: “China values the protection of intellectual property rights,” he affirmed.
China’s Foreign Ministry spokesperson, Guo Jiakun, labeled the U.S. claims as baseless allegations aimed at undermining China’s AI achievements.
“China firmly opposes this. We urge the U.S. to acknowledge facts, abandon biases, cease obstructing China’s technological progress, and foster scientific and technological exchanges and cooperation,” he added.
Additionally, Kratsios’ memorandum coincided with a unanimous, bipartisan endorsement from the House Foreign Affairs Committee for legislation that seeks to establish a methodology for identifying foreign actors who extract crucial technical features from closed-source, U.S.-owned AI models, with punitive measures, including potential sanctions.
According to Representative Bill Huizenga, a Republican from Michigan and bill sponsor, “Model extraction attacks represent the latest frontier of Chinese economic coercion and intellectual property theft.”
He stressed, “It is essential that we thwart China’s attempts to pilfer these technological breakthroughs.”
Last year, the Chinese startup DeepSeek set U.S. markets abuzz following the unveiling of a large language model capable of competing with American AI titans at a fraction of the cost.
David Sacks, who at the time served as adviser for AI and cryptocurrency under President Trump, indicated that DeepSeek may have drawn from U.S. models. He stated, “There’s substantial evidence that DeepSeek distilled knowledge from OpenAI’s models.”
In February, OpenAI, the minds behind ChatGPT, echoed similar contentions, advocating that China should not be permitted to champion “autocratic AI” by “appropriating and repackaging American innovation.”
Furthermore, Anthropic, developer of the Claude chatbot, accused DeepSeek and two other China-based AI institutes of orchestrating efforts to “illicitly extract Claude’s capabilities” to enhance their own models through a distillation technique that exploits training on the outputs of superior models.
While such distillation can be an acceptable method for training AI systems, it becomes problematic when competitors leverage it to gain powerful capabilities in a fraction of the time and cost that independent development would entail.
Nonetheless, this intricate game may not be one-sided. San Francisco’s Anysphere, creator of the popular coding tool Cursor, recently acknowledged that its latest product was derived from an open-source model fashioned by Chinese firm Moonshot AI, developers of the chatbot Kimi.
Kyle Chan, a fellow at the Brookings Institution specializing in Chinese technological advancements, remarked that discerning unauthorized distillation from legitimate data requests will be akin to “searching for needles in a vast haystack.”
He asserted, however, that enhanced information sharing and coordination among U.S. AI laboratories, coupled with federal initiatives, could bolster anti-distillation efforts.
The efficacy and potential trajectory of the House bill remain uncertain. Yet, Chan posits that Trump may be hesitant to escalate tensions with Chinese President Xi Jinping ahead of a scheduled state visit to Beijing in mid-May.
Source link: 1news.co.nz.
