Garry Tan says "I would do nothing" about China's AI distillation and urges the industry to focus on current AI risks instead of doomsday-style extinction fears

If you are concerned about Chinese competitors copying AI models, the primary accelerator for startups says there is no immediate action to take.

Garry Tan, CEO of Y Combinator, stated he would "do nothing" regarding accusations that Chinese companies are distilling AI models from OpenAI and Anthropic. He argues that the practice of distillation, where smaller models are trained on the outputs of larger ones, is not a primary concern for regulators. Tan believes the focus should instead be on current AI risks such as cybersecurity threats and bioweapon research, rather than speculative doomsday scenarios. He also highlighted that data used for AI training may be subject to copyright, a matter currently under legal scrutiny in various lawsuits. Tan advocates for a balanced approach, ensuring open-weight models offer freedom and access while frontier models maintain a price premium to sustain their business models. He believes significant societal impacts like job loss are decades away, suggesting a transition towards automation of routine tasks and increased focus on creative work.

Tan's stance suggests a market perspective that views model distillation as a natural progression and a driver of innovation, rather than a threat to be policed. This approach could embolden open-source AI development globally, potentially accelerating the democratization of advanced AI capabilities. It also signals a potential divergence in regulatory priorities between major AI developers seeking to protect their intellectual property and a segment of the tech ecosystem that favors open access and rapid iteration.

The emphasis on current, tangible risks like cybersecurity and misuse for bioweapons suggests a shift in how AI safety is being framed, moving away from abstract existential threats towards immediate practical concerns. This reorientation could influence where research funding and policy attention are directed, potentially impacting the development of AI safety protocols and international cooperation on AI governance, prioritizing immediate threats over long-term speculative ones.