Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
Tech Wire Asia is part of the TechForge Publications series
TechForge
AI & Intelligence, Compliance, Regulations & Governance, Generative AI & LLMs, Government Initiatives & Policy, Sovereignty & Privacy
Muhammad Zulhusni
February 13, 2026
Share this story:
Tags:
Categories::
Competition between US and Chinese AI developers is moving beyond model performance and into how those models are trained. A growing concern in Washington and Silicon Valley is whether advanced systems can be copied through indirect means, narrowing the gap between companies that invest heavily in infrastructure and those that build on their output.
According to reporting by Bloomberg News, OpenAI has warned US lawmakers that Chinese AI firm DeepSeek is using increasingly sophisticated techniques to extract results from American models to help train competing systems. The warning appeared in a memo OpenAI sent to the House Select Committee on China, where the company described what it sees as attempts to bypass safeguards and reuse its model outputs.
At the centre of the dispute is a training practice known as distillation. Instead of accessing a rival model’s code or internal data, a developer feeds it prompts and uses the responses to train another model. In theory, this allows a system to absorb patterns and behaviour from a more advanced model without direct access to its architecture. OpenAI told lawmakers it has detected “new, obfuscated methods” designed to mask when such activity is taking place, Bloomberg reported.
The company characterised these efforts as attempts to “free-ride on the capabilities developed by OpenAI and other US frontier labs.” Distillation itself is not illegal or new, but OpenAI argues that large-scale automated extraction can undermine the economic model behind frontier AI research. Training state-of-the-art systems requires major spending on computing infrastructure, specialised chips and engineering talent. If competitors can replicate behaviour through output harvesting, the return on that investment becomes harder to protect.
This tension carries commercial weight. Many Chinese AI systems, including DeepSeek’s R1 chatbot, operate without subscription fees. Bloomberg noted that widespread distillation could give low-cost competitors a shortcut to parity with US firms that charge for access. For American developers, the concern is not only lost revenue but a broader erosion of competitive advantage in a sector tied to national technology leadership.
OpenAI’s memo also raised safety questions. The company warned lawmakers that when capabilities are reproduced through distillation, embedded safeguards may not transfer cleanly. That gap could increase the risk of misuse in areas such as chemistry or biology, where controlled outputs are part of model design.
Bloomberg also reported that OpenAI highlighted examples where DeepSeek’s chatbot appeared to censor responses on topics sensitive to the Chinese government, including Taiwan and the Tiananmen Square crackdown.
Political reaction has been sharp. Representative John Moolenaar, chair of the House China committee, framed the issue in strategic terms. “This is part of the CCP’s playbook: steal, copy, and kill,” he said. “Chinese companies will continue to distill and exploit American AI models to their advantage, just like when they ripped off OpenAI to build DeepSeek.” Bloomberg reported that OpenAI declined further comment, while representatives for DeepSeek and the Chinese embassy did not immediately respond to requests.
The memo suggests OpenAI’s defensive measures have not fully contained the practice. According to Bloomberg, the company believes accounts linked to DeepSeek employees tried to route access through third-party services to hide their origin. OpenAI also cited code designed to pull model outputs in automated ways, alongside what it described as networks of unauthorised resellers offering indirect access.
Concerns about distillation intersect with a wider debate over hardware access. White House AI adviser David Sacks said last year that DeepSeek was “squeezing more juice” out of older chips and cited “substantial evidence that what DeepSeek did here is they distilled the knowledge out of OpenAI’s models.”
The hardware dimension remains sensitive because advanced chips are central to training competitive systems. Bloomberg reported that late last year export rules were eased to allow Nvidia to sell H200 processors to certain markets, even as policymakers continued to debate how chip controls affect AI competition.
Questions around DeepSeek’s infrastructure have also drawn scrutiny. Bloomberg previously reported that US authorities examined whether the company bypassed export controls to obtain processors through intermediaries. Documents reviewed by the House China committee indicated that Nvidia provided technical support during development of DeepSeek’s R1 model, which required millions of GPU hours to train.
Representative Michael McCaul, who formerly oversaw export control policy in Congress, warned about the strategic implications. “DeepSeek should have been a wake-up call about the dangers of selling advanced semiconductor chips to the CCP,” he said. “Using less powerful Nvidia chips, China developed the most advanced open-source models on the planet. I shudder to think of what they might do with more advanced hardware like the H200 chips.”
For enterprises watching the AI sector, the dispute highlights how competitive boundaries are shifting. Model development is no longer defined only by raw training power or algorithm design, but by how effectively companies can protect output, enforce access rules and manage cross-border technology flows.
Bloomberg’s reporting shows that distillation has moved from a technical curiosity to a focal point in the geopolitical contest over AI leadership — one that blends business strategy, policy enforcement and platform security.
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.
AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
Muhammad Zulhusni
Muhammad Zulhusni
February 13, 2026
February 12, 2026
February 12, 2026
February 11, 2026
Subscribe now to get all our premium content and latest tech news delivered straight to your inbox
China, South Korea, Taiwan
Digital Transformation
Uncategorized
AI & Intelligence, Software & Applications
Asia Pacific Focus
February 12, 2026
Asia Pacific Focus
February 11, 2026
Australasia
February 11, 2026
All our premium content and latest tech news delivered straight to your inbox
TechWire Asia is part of TechForge
All our premium content and latest tech news delivered straight to your inbox