How to Make Artificial Intelligence Sustainable

Building AI in an environmentally responsible way may seem like an impossible goal, especially as major tech companies that once pledged to cut emissions are now racing to expand massive data centers fueled by fossil fuels. The urgency to develop AI at any cost has only intensified under the Trump administration, which is simultaneously rolling back environmental safeguards.

Yet despite these challenges, Sasha Luccioni, a researcher focused on AI sustainability, believes that demand for transparency in AI has never been higher from customers, both businesses and individuals alike.

Over the past four years at Hugging Face, an AI company, Luccioni has emerged as a leading voice pushing for greater openness about AI’s emissions and environmental footprint. She pioneered a leaderboard that tracks the energy efficiency of open-source AI models and has been a vocal critic of major AI firms, accusing them of deliberately hiding energy and sustainability data from the public.

Now, she is launching Sustainable AI Group, a new venture co-founded with former Salesforce sustainability chief Boris Gamazaychikov. The organization will help companies answer a crucial question: “What are the levers that we can play with in order to make agents slightly less bad?” Luccioni is also eager to explore the energy demands of various AI tools, such as speech-to-text translation or photo-to-video generation, an area she says has received far too little research attention.

In an exclusive interview with WIRED, Luccioni discussed the growing appetite for sustainable AI and what she wants from Big Tech. The conversation has been edited for length and clarity.

WIRED: I hear from many individuals worried about AI and the environment, but less from companies. What are businesses working with AI telling you?

Sasha Luccioni: First, they face mounting pressure from employees and boards. Directors are saying, “You need to quantify this.” Employees are asking, “You’re forcing us to use Copilot,how does that affect our ESG goals?”

For most companies, AI is now central to their offerings. So they must understand the risks. They need to know where models are running. They can’t keep using models without knowing the location of data centers or the grid they’re connected to. They have to account for supply chain emissions, transportation emissions, and all those factors.

It’s not about abandoning AI. I think we’re past that. It’s about choosing the right models, for example, or signaling that energy source matters, so customers are willing to pay a bit more for data centers powered by renewable energy. There are ways to do this,it’s a matter of finding the believers in the right places.

I’d imagine that for global companies, the sustainability landscape differs from the US. The US government may not care, but others certainly do.

In Europe, the EU AI Act has included sustainability as a significant component from the start. They added several clauses, and now the first reporting initiatives are emerging.

Even Asia is pushing for more transparency. The International Energy Agency has been publishing reports on AI and energy use. I spoke with them, and they noted that other countries realize the IEA gets its numbers from the countries themselves,but those countries lack specific data on data centers. Without those numbers, they can’t make informed future decisions, like “We need X capacity in the next five years.” Some countries have already started pushing back against data center builders.