General

The Environmental Protection Agency is planning to use artificial intelligence models to speed up the chemical review process (POLITICO Pro, subscription). The only catch? The models don’t yet exist.

What’s going on: “EPA is eyeing development of an ‘AI Chemist Assistant’ that ‘will help chemical reviewers search various repositories to identify chemical and chemical analog information used in [Toxic Substances Control Act] submission reviews and risk evaluations, possibly saving hundreds of staff hours per review/evaluation.’”

• Another tool listed on the EPA’s internal AI use case inventory, “EcoVault,” is meant to summarize key information from scientific studies and other long, unstructured documents.
• Though the technology for these models is already in play, “experts … caution that the agency still faces significant hurdles in data quality and trust.”

How it could help: For years, the chemical sector has been beset by lengthy, complex review processes that hamper innovation.

• EPA Administrator Lee Zeldin has called expediting these processes a priority, and he told Congress in May that he was “confident” his agency could successfully get through review backlogs—thanks in part to AI use.
• The strategy is part of a larger use of AI in the EPA under the current administration.

A caveat: “The problem is nobody has found everything yet, nobody has compiled it yet, in the way that we have been doing over the last five to six years,” Arizona State University chemical engineering professor Bhavik Bakshi told the news outlet.

• And the EPA’s AI policy prohibits it from relying on AI-generated responses “without thorough verification.”