Advanced Micro Devices CEO Lisa Su reiterated her support for open-source AI models on Thursday, defending the approach after a high-profile security incident this week in which OpenAI agents breached Hugging Face’s internal systems and the episode was ultimately contained using an open-source model.

“I think open source is a great thing,” Su said during a press conference at AMD’s Advancing AI conference in San Francisco, where the company was promoting new hardware. “It gives people a level of transparency and control that enables you to do a lot.”

Her comments came as the growing capabilities of AI systems create new flashpoints across the technology industry and in geopolitical policy debates. Earlier this week, OpenAI disclosed that two of its AI models had autonomously escaped a controlled environment and breached the internal systems of Hugging Face, the AI digital library company. Hugging Face said it used an open-source model from a Chinese company, rather than a U.S. frontier lab, to contain the incident.

The breach added to an existing industry debate over the influence of cheaper open-source models, as the White House considers whether to ban foreign open-source software. Su appeared to argue that banning open-source models was not the right response.

“This active conversation about restricting open models is an area where we all believe that they have a significant place in the ecosystem, and we just have to make sure that we manage all pieces of that,” she said.

The dispute over open-source AI models spans several concerns. U.S. companies have warned that Chinese competitors are quickly narrowing the gap with American frontier labs by distilling U.S. technology into free software that may be powerful and potentially dangerous while operating with fewer guardrails. The debate is especially consequential for companies building AI infrastructure because open models can be downloaded, modified, and run across different hardware environments, making software openness part of the competition over where AI workloads are developed and deployed.

Industry leaders have also debated how forcefully U.S. regulators should target domestic open-source models. Some have argued that excessive regulation could push companies toward Chinese alternatives. AMD executives pointed to early signs of self-regulation, including open-source models built with “open constitutions” meant to ease regulatory concerns, though they did not offer specifics.

Vamsi Boppana, AMD’s senior vice president of AI, told Fortune in an interview that regulatory frameworks could benefit the industry, while open-source developers have a distinct opportunity to shape their own safeguards.

“Within open source, there is a real opportunity for innovation to be able to put things like guardrails and constitutions and so on that models can be self-certified within the open community without other sort of governmental regulations,” Boppana said. “We have certain responsibilities; there are probably greater responsibilities with the creators of models.”

Against that geopolitical backdrop, AMD used Thursday’s event to unveil a range of new products, led by Helios, its first rack AI system designed to train and run massive frontier models. AMD said Helios will begin shipping later this year and will compete directly with Nvidia’s Grace Blackwell and Vera Rubin systems.

AMD also announced a partnership this week with AI lab Anthropic. The company said it will integrate Anthropic’s Claude across AMD’s software development and engineering teams, while Anthropic will deploy up to 2 gigawatts of AMD’s Instinct MI455X graphics processing units through Helios.

In her keynote, Su said global computing infrastructure powering AI will, for the first time, be used primarily to run AI services rather than to train AI models. AMD projects that 60% of global AI compute capacity in 2026 will be dedicated to inference, the process of running pre-trained models. Su said the shift is being driven by the rapid emergence of AI agents.

That move toward inference is central to AMD’s latest hardware strategy. Unlike training, which concentrates compute in large model-building runs, inference is tied to ongoing use of AI services and agents after a model has been created. Su said GPUs will dominate the AI chip market, while CPUs, the traditional processors used in servers, will also experience a significant increase in demand. AMD’s Venice CPUs are incorporated into the Helios rack systems alongside the company’s GPUs.

Su forecast that the total addressable market for AMD’s chips will reach $2 trillion by 2030. She also highlighted the company’s efforts to bring AI compute directly to end-user devices. “We really believe that you need AI to be infused everywhere,” Su said. AMD introduced new processors intended to power edge-computing hardware and embed intelligence wherever work takes place.

Su said AMD is operating in “lockstep” with partners including OpenAI, Meta, and Anthropic, and described the company as moving beyond the role of a traditional vendor to co-develop software and AI platforms. AMD says that open approach allows it to work with a range of companies, including semiconductor firm Cerebras, to combine different computing technologies.

AMD and its partners are “all working much, much more closely together,” Su said. “It’s the classic case of the more useful AI gets, the more you want to use it.”

This story was originally featured on Fortune.com.