Pharmaceutical companies navigating the regulatory landscape around artificial intelligence may be drawing overly cautious conclusions from federal guidance, according to a former regulator who has since moved into the biopharma industry.
Tala Fakhouri, who previously worked on AI regulation at the FDA, has argued that the industry's reading of the agency's AI guidance is more restrictive than the framework actually requires. Speaking to STAT News, Fakhouri offered a perspective that is relatively rare: an insider view shaped by experience on both sides of the regulatory divide.
Companies are being too conservative in how they interpret FDA's AI guidance, but the agency can do more to help, too.
The observation carries particular weight given Fakhouri's background. Having helped shape AI-related regulatory thinking from within the FDA before transitioning to industry, her position allows for a dual critique — one directed at pharmaceutical firms for their overcautious stance, and another directed at the agency for not doing enough to reduce ambiguity.
A Tension Between Caution and Progress
The broader context is one of rapid technological change colliding with a regulatory environment that has struggled to keep pace. Artificial intelligence applications in drug development, clinical trial design, and safety monitoring have expanded considerably in recent years, prompting the FDA to issue guidance intended to help companies understand what is and is not permissible.
Yet guidance documents, by their nature, leave room for interpretation. When the stakes involve drug approvals and patient safety, companies tend to err on the side of restraint — sometimes to a degree that, according to Fakhouri, goes beyond what regulators actually intend.
This dynamic is not unique to AI. Pharmaceutical firms have long been known to apply conservative interpretations to ambiguous regulatory signals, particularly in areas where enforcement history is limited. With AI, however, the pace of development means that overly cautious readings could slow the adoption of tools that might otherwise improve drug discovery timelines or enhance the detection of adverse events.
Responsibility on Both Sides
Fakhouri's critique is notably balanced. While she places some of the interpretive burden on industry, she also suggests the FDA has room to improve how it communicates its expectations. Clearer guidance, more direct engagement with stakeholders, and additional explanatory materials could all help reduce the uncertainty that drives conservative interpretations in the first place.
This dual accountability framing is significant. It resists the tendency to assign blame exclusively to either regulators or regulated entities, instead framing the disconnect as a communication failure that both parties have a role in addressing.
The Regulator-to-Industry Pipeline
Fakhouri's career trajectory — from federal regulator to biopharma — also raises broader questions about how expertise moves between the public and private sectors. Former regulators who join industry often bring nuanced understanding of how agencies think, which can help companies engage more effectively with oversight bodies. At the same time, such transitions attract scrutiny over potential conflicts of interest.
In this instance, Fakhouri appears to be using her dual vantage point to push for greater clarity and more productive dialogue between the FDA and the companies it oversees, rather than to advocate for deregulation or looser standards.
What This Means for AI in Drug Development
The stakes of getting AI regulation right in biopharma are considerable. Artificial intelligence tools are increasingly embedded in processes ranging from target identification and molecular screening to post-market surveillance. How companies interpret regulatory expectations will shape how aggressively — or cautiously — they deploy these technologies.
If Fakhouri's assessment is accurate, a significant portion of the hesitancy currently observed in the sector may stem not from genuine regulatory prohibition but from interpretive conservatism that the FDA has not yet done enough to correct. That gap, she suggests, is one that both industry and the agency have an interest in closing.
The conversation around AI governance in pharmaceuticals is still evolving, and perspectives from those who have operated within the regulatory apparatus itself are likely to remain influential as the FDA continues refining its approach.