AI has started producing drug-like antibodies. When will it revolutionize biopharmaceuticals?

Someday, perhaps within the next year or two, a company will claim to have brought the first artificial intelligence-designed antibodies to the clinic. But the industry is divided on what “AI design” actually means and how close we are to technology that can truly design medicines.

What exactly does it mean that antibodies are designed by AI? Antibody and protein AI researchers interviewed by STAT are of two schools of thought. For some, if a computer designs a basic antibody sequence that scientists then fine-tune to create a clinical candidate, it is considered “designed by AI.” Others say that to be truly engineered by AI, an antibody would need to be ready to go directly from a computer to a clinical setting, with no further experimental work required, a much higher hurdle to clear.

In 2025, researchers revealed that AI can meet the first simple definition of creating “AI-designed” antibodies. But even though some startups claim to be producing clinically-ready antibodies and investors are pouring more money into AI-native biotechs at higher valuations compared to traditional biotech startups, even pharmaceutical and antibody experts who have embraced AI find it hard to believe that AI models of de novo protein design can match or outperform traditional technologies.