Y Combinator CEO Has A Surprising Message For OpenAI And Anthropic: ‘I Would Do Nothing’

OpenAI and Anthropic are raising concerns about Chinese AI companies allegedly using their models to build competing systems. But Y Combinator CEO Garry Tan has a surprisingly simple response: do nothing.

Tan said he would not support regulators stepping in to stop AI model distillation, a technique that allows developers to use the responses of a powerful AI model to help train another model.

“I would do nothing” about distillation, Tan said while speaking to CNBC during Y Combinator’s annual Demo Day.

His comments come at a time when the debate around AI competition, intellectual property and national security is becoming increasingly heated. OpenAI and Anthropic have both alleged that Chinese AI developers have used their models to improve competing

systems, while US officials have also raised concerns about the technology.

Why Are OpenAI And Anthropic Worried About AI Distillation?

AI model distillation is not a new concept. In simple terms, it involves using the answers or behaviour of a more powerful AI model to help develop a smaller or different model.

The technique can be a legitimate part of AI research and development. The controversy begins when developers allegedly extract information from another company’s model without permission or in ways that violate its terms of service.

Anthropic has accused Chinese AI companies including Moonshot AI, DeepSeek and MiniMax of distilling its models. OpenAI has also alleged that DeepSeek’s V3 and R1 systems were distilled from its GPT-4 and GPT-4o models.

It is important to note that these are allegations made by the companies involved and do not, by themselves, establish that the alleged activity occurred.

The issue has also attracted attention from US government agencies. The National Security Agency, Cybersecurity and Infrastructure Security Agency and Federal Bureau of Investigation have warned about the potential risks associated with AI model distillation.

Garry Tan Says Regulators Should Look At The Bigger AI Picture

Tan does not believe stopping distillation should be the priority. Instead, he suggested that regulators should focus on creating a workable balance between powerful frontier AI models and open-weight models.

Open-weight AI models make their trained model weights available for others to use, adapt or build upon. Frontier models, meanwhile, are the most advanced systems developed by companies such as OpenAI and Anthropic.

Tan argued that both approaches have a role to play. Open-weight models can give developers greater freedom and access, while companies building frontier systems need enough commercial advantage to continue investing billions of dollars in AI research and infrastructure.

“We could argue that there should be an American distillation regime,” Tan said.

He described the regulatory challenge as a “tightrope”, but suggested that getting the balance right could ultimately produce a better outcome for the AI industry.

His broader argument is that regulators should be careful about creating rules that could unintentionally limit AI development or reduce access to useful technology.

Tan Also Questions Some Of The Bigger AI Safety Fears

Tan also used the discussion to argue that the AI safety debate should remain focused on risks that can be demonstrated rather than scenarios that may still belong to science fiction.

“We need to be focused on science fact, not science fiction,” he said.

For Tan, cybersecurity is one of the more immediate concerns. Instead of concentrating only on hypothetical future scenarios, he believes the industry should think about what happens if AI systems are actually used in a coordinated attack on critical infrastructure.

His comments come as AI companies increasingly confront questions about how their systems could be misused. Anthropic recently said it had blocked Claude access in five cases involving researchers from unspecified foreign countries who were using the AI model for work involving dangerous pathogens.

Anthropic said it was concerned that the researchers could potentially have been using Claude in connection with biological weapons development.

What About AI Taking People’s Jobs?

Tan also offered a more measured view of the economic impact of AI. Concerns about AI replacing workers have become one of the biggest debates surrounding the technology. As companies automate more tasks, there are fears that large numbers of jobs could disappear.

Tan believes the transformation will be slower than some predictions suggest. He expects businesses to gradually automate repetitive work while employees spend more time on creative and higher-value tasks. “It will take decades for this to actually percolate into society, and that’s not a bad thing,” he said.