Block’s humanlike voice model draws praise and scam concerns

Jack Dorsey’s payments company Block has unveiled Dessa, a voice AI trained entirely on human speech that processes audio directly rather than converting it into text. The model can listen while speaking, handle interruptions and respond to changes in tone. Company demonstrations feature laughter and seemingly emotional reactions, with Block saying the conversations were recorded live without anyone directing the responses.

Block says Dessa was trained from scratch using a single Nvidia GB200 NVL72 computing system and can run locally on a MacBook Pro equipped with an M4 Max chip or better. It remains a research preview, with no disclosed model size, benchmark results or public release date. The available audio samples are company demonstrations, not independent tests.

Social media reactions ranged from praise for its natural conversation to warnings about deception and more convincing scam calls. Binance founder Changpeng “CZ” Zhao suggested the technology could help his GiggleAcademy education project teach children through conversation. Filmmaker Justine Bateman questioned whether humanlike voices could encourage people to lower their guard.

Block is offering early access to a small group of AI safety researchers to assess overlooked risks. The company also sees potential for speech-based systems to recognize distress, pain and other emotional cues that text transcripts can miss.