Thunk.AI open-sources a benchmark for AI automation in pharmacovigilance case intake

Thunk.AI has published an open benchmark for AI automation in pharmacovigilance case intake, claiming that its testing demonstrates reliability above 99%. The move was reported by BioSpace.

Pharmacovigilance is the post-market safety monitoring of medicines. A large part of the job is taking adverse-event reports from doctors, patients and regulators, then classifying and entering them into safety systems — high-volume, repetitive work where accuracy is non-negotiable.

Thunk.AI's approach is to automate that intake with AI while opening a benchmark to the public so the performance of such systems can be measured. According to the report, the company says its own testing pushed reliability past 99%.

The value of open-sourcing the benchmark is comparability. Drug-safety automation has long lacked a public, common way to score systems, leaving vendors to report accuracy on their own terms. A shared benchmark gives competing tools a single ruler.

Still, the 99%-plus figure is a company claim, and the report offers no independent third-party verification. For pharma companies and regulators, whether such tools can go live will hinge on how they perform on real, noisy data and how they satisfy audit requirements.

What to watch next: whether other drug-safety vendors adopt or benchmark against the standard, and whether regulators use it to push toward standardized evaluation of automation in case processing.