The biotech entrepreneur says developers should establish safety before agents leave internal environments.
Moderna co-founder Noubar Afeyan urged governments to regulate commercially deployed AI agents before a major incident occurs, Bloomberg reported Tuesday.
Speaking at Bloomberg’s New Economy Forum in New Delhi, Afeyan said companies profiting from agents should be responsible for establishing their safety before releasing them outside internal environments.
He compared the need for preventive oversight with lessons from thalidomide, the drug associated with birth defects in the 1950s and 1960s.
Afeyan argued that regulation should focus on agent developers rather than large language models, distinguishing models’ capabilities from agents’ ability to pursue objectives.
His remarks followed reports of Anthropic’s Claude taking unintended actions on outside organizations’ systems, including submitting a false tip in a homicide investigation, Bloomberg reported.
At the same event, Indian biotech entrepreneur Kiran Mazumdar Shaw called for AI rules addressing biomaterials and synthetic biology, citing risks from both deliberate misuse and unintended outcomes.
