FDA doubles down on its push into AI

The FDA has revealed the latest element in its adoption of AI, an agentic AI platform that agency staffers can use to streamline complex tasks.

According to the US regulator, agentic AI – autonomous systems capable of independent decision-making – can now be used to make "more complex AI workflows […] to assist with multistep tasks."

That includes meeting management, pre-market reviews, review validation, post-market surveillance, inspections, and compliance and administrative functions, according to the agency, which also told Stat that the rollout is currently exploratory and the agentic systems will not make regulatory decisions or replace human judgment.

The FDA's announcement further stressed that the models "do not train on input data nor any data submitted by regulated industry, safeguarding the sensitive research and data handled by FDA staff."

The FDA has made no secret of its intention to harness AI across its operation, charting that in a guidance document (PDF) at the start of this year, and the process has accelerated under new Commissioner Marty Makary.

The new agentic platform follows the launch of a generative AI (GenAI) system called Elsa earlier this year to speed up tasks like clinical protocol reviews, scientific evaluations, and selecting sites for inspection.

The drive to roll out AI comes after sweeping job losses at the FDA that resulted from the cost-cutting drive implemented by Elon Musk and the now-defunct Department of Government Efficiency (DOGE), dissolved with eight months left on its contract and with savings only a tiny fraction of the initial claim of $2 trillion in its first year.

Makary did not make mention of FDA resourcing in his announcement on the agentic AI tool, but said the FDA is "diligently expanding our use of AI to put the best possible tools in the hands of our reviewers, scientists, and investigators."

He also said that the Elsa tool is now being used voluntarily by more than 70% of agency staff, citing internal data, and said that FDA teams have been modifying the large language model (LLM) so it integrates better with workflows.

The intention is to follow a similar path with the agentic AI, and the FDA is launching a two-month challenge for staff to build solutions and demonstrate them at a Scientific Computing Day in January.

"FDA's talented reviewers have been creative and proactive in deploying AI capabilities – agentic AI will give them a powerful tool to streamline their work and help them ensure the safety and efficacy of regulated products," commented the company's first chief AI officer, Jeremy Walsh, who took up his role in May.

Image by Gerd Altmann from Pixabay