Salesforce expands disclosures about AI environmental impact

This is corporate software news, not a home energy story, so it has little bearing on your own house. Salesforce, the business software company, has started adding energy use and carbon emissions data to the "model cards" it publishes for some of its artificial intelligence tools. Model cards are fact sheets that explain how an AI model was built, how it performs, and where it might be used. Salesforce is now including electricity consumption and emissions figures alongside that information, something few other companies in the industry do.

The change applies first to models Salesforce trained itself, including tools used to match names and accounts in its software. The figures come from the AI Energy Score, a measurement system built by Hugging Face, an open-source software company, in a project started by former Salesforce sustainability executive Boris Gamazaychikov and computer scientist Sasha Luccioni. That system looks at the computing hardware, data center location, and run times behind common AI tasks like image generation or speech recognition.

Salesforce says it plans to expand this reporting to more of its AI models over time, as it gets easier to measure the energy used when AI is actually put to work, not just when it is trained. For homeowners, this is a look at how AI companies are starting to disclose environmental costs of their products, not a change to any home energy program or rebate.