They Taught Themselves to Hack

In March 2026, researchers at Irregular, a frontier AI security lab backed by Sequoia Capital, published findings that should unsettle anyone who has ever typed a password, visited a doctor, or sent a...