AI can now sign its own artwork: scientists watermark lab-made proteins without breaking them

As AI starts designing brand-new proteins for medicine, a fair question follows: how do we know where one came from? Google DeepMind researchers have published a method in Nature that hides an invisible signature inside AI-designed proteins and predicted structures — and lab tests on hundreds of designs, checked by Carterra, show the marked proteins still grab their targets just as well as unmarked ones. A small signature with big uses for safety and scientific honesty.