🔍 Read the full analysis: Why Block’s Dessa’s Human-Like Voice Isn’t Evidence Of A Person on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Block introduced Dessa, a speech-to-speech research model whose company-selected demos include laughter, emotional expression and interruptions. Those recordings do not establish that the system is a person, that it reliably performs as described, or that it can authenticate a speaker; outside researchers have not tested it.
Block announced Dessa, a research model that listens and speaks in real time, and released recordings in which it laughs, responds to emotion and interrupts naturally. The demonstrations show what Block says its system can do; they do not establish that Dessa is a person or independently verify its abilities, because the model has not been publicly released or tested by outside researchers.
Block, the company behind Square and Cash App, described Dessa on a project page published October 23. According to Block, the model was trained from scratch on recorded human conversation, without text or pretrained models. It handles speech in both directions at once, which Block calls full-duplex conversation, and is intended to run faster than real time on a MacBook Pro M4 Max or better. The company has not provided a specific latency figure.
Block’s recordings feature the system laughing during a sentence, reacting to a surprise, and helping a speaker recall a film title. The company says the clips are live conversations and that nobody steered the system. These remain company-selected demonstrations, not independent evaluations. The Verge reported that some exchanges could feel awkward or uncanny, while TechCrunch said it was unclear whether emotional delivery was guided by tags. CoinDesk reported that Block had not disclosed the model’s size, benchmark results or public release date.
As of October 26, Block had published no technical paper, model card, license or public weights or API. Access was through a waitlist, with early access for a small group of AI safety researchers, according to the source report. A convincing voice is evidence of a system producing convincing audio; it is not evidence that the speaker has consciousness, feelings, a human identity or a particular person’s consent.
Infographic 01 · Dessa fact sheet
What Block claims, what was heard, what is missing.
Every capability below comes from Block or its selected recordings. Nobody outside Block has run the model yet.
| Topic | What we know | Status |
|---|---|---|
| Training data | Human conversation only: “no text, no pretrained models” | Block’s claim |
| Conversation | Full-duplex: listens while it talks, backchannels, cuts in | Block’s claim |
| Emotion | Hears and speaks laughter, sarcasm, sighs; unclear if steered by tags (TechCrunch) | Block’s claim |
| Speed | “Faster than real time”; no latency figure | Block’s claim |
| Hardware | Trained on one NVIDIA GB200 NVL72; runs on a MacBook Pro M4 Max or better | Block’s claim |
| Demos | Some chats “a little awkward and uncanny”, more natural than GPT-Live-1 (The Verge) | reporters heard it |
| Size, benchmarks, release date | Not disclosed (CoinDesk) | not disclosed |
| Paper, model card, license, weights | None found as of 26 Oct 2026 | not disclosed |
| Languages | Not listed; Block calls learning new languages an open question | not disclosed |
| Voice cloning safeguards | No statement on watermarking, consent checks or imitating a specific person | not disclosed |
Who can use it
research preview
- Public: a waitlist; “launch links are coming soon”
- Early access: “a small group of AI safety researchers”, institutional email required
- Business use: none announced; Square already uses voice AI for ordering (TechCrunch)
Who built it
from Block’s page
- A “(very) small AI research lab at Block” reporting to Jack Dorsey
- Founded by former members of Dessa, the Toronto startup Square acquired in 2020 (CoinDesk)
- Lead Ragavan Thurairatnam: “a GPT3 moment for speech”
Sources: Block, Dessa project page (23 Oct 2026); Ragavan Thurairatnam on X (23 Oct 2026); The Verge (23 Oct 2026); TechCrunch (24 Oct 2026); CoinDesk (24 Oct 2026).
Powered by Thorsten Meyer AI
Why Voice Alone Cannot Verify Identity
The announcement matters because natural conversational behavior can make synthetic speech harder to recognize during a call. Laughter, pauses and interruptions may make an interaction feel familiar, but they do not prove who—or what—is speaking. A human-sounding voice should not be treated as an authentication credential.
The FBI’s 2025 Internet Crime Report, published in April 2026, counted 22,364 complaints mentioning AI and $893 million in reported losses. It also reported more than $5 million in losses from distress scams using cloned voices to imitate loved ones. These figures cover complaints and losses attributed to AI in that report; they do not establish that Dessa was used in any scam. The risk is broader: voice-cloning techniques already exist, and better conversational realism could remove a cue some callers use to spot impersonation.
For businesses, a familiar-sounding caller should not by itself authorize a transfer or account change. Independent callback procedures and written approval can add checks that a voice cannot supply. For families, a pre-agreed code word and calling back on a known number can help verify an urgent request. Those steps address impersonation risk generally; the source material does not say that Dessa itself is being used for fraud.
Products Worth Considering
Brother TD-2020A 2‑Inch Direct Thermal Desktop Label Printer with USB
HP Sprocket Panorama Portable Color Photo and Label Printer
Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo
As an affiliate, we earn on qualifying purchases.
What Block Has—and Has Not—Shown
Block’s project page says Dessa can hear and produce emotion, sarcasm, laughter and vocal cues such as a frustrated sigh. Co-founder Jack Dorsey described the experience as feeling like a conversation with a person, pointing to its emotion, nuance, pauses and cadence. Dessa lead Ragavan Thurairatnam called the project “a GPT3 moment for speech,” according to the source report. These are attributed descriptions of the system, not results from published benchmarks.
The technical claims are also incomplete. Block says the model learned from “hundreds of times fewer words” than large language models require and was trained using one NVIDIA GB200 NVL72. It has not published the underlying training details or independently reproducible results in the material available as of October 26. No outside party has run the model, and there is no confirmed list of supported languages. Block itself has described how efficiently Dessa can learn new languages as an open question.
The distinction between voice quality and personhood is central. A model can generate emotionally expressive speech without having emotions, and a recording cannot establish that it understands a conversation as a human does. The Decoder made that distinction in its coverage, saying that sounding emotional is not the same as being emotional or intelligent. The demonstrations can support an assessment of what those particular clips sound like; they cannot answer questions about subjective experience or general reliability.
“Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.”
— Jack Dorsey, Block co-founder
Unanswered Questions About Dessa
Important details remain unverified: model size, benchmark performance, latency measurements and the training data have not been fully documented in the material available. Block has not announced public weights or an API, and no public release date was reported as of October 26. The demos do not establish how the system performs across long conversations, different speakers or languages, or under independent testing.
It is also unclear whether the model can imitate a specific person, what safeguards or consent checks it uses, and whether its output is watermarked. Block says Dessa chooses its own voice and personality, but that statement does not resolve those questions. The source report also notes uncertainty about whether emotional delivery in the demos may be influenced by tags. No evidence in the supplied material establishes that the model has feelings or is conscious.
Independent Testing and Release
The next meaningful evidence would be a technical paper, model card or controlled access that lets researchers evaluate the system. Those materials could clarify its performance, limitations, supported languages, safeguards and whether the demo behavior holds up beyond selected clips. Block had not announced a public release date as of October 26, 2026; its waitlist and limited access for safety researchers were the reported routes to access.
Until independent findings are available, readers should treat the recordings as demonstrations rather than proof of general capability or personhood. Companies can verify sensitive requests through a separate channel instead of relying on a familiar voice. Families can agree on a code word and call back using a number already known to them. The FBI figures cited here describe reported 2025 losses, not a forecast of future fraud or evidence that Dessa caused those losses.
Key Questions
Does Dessa prove that an AI can feel emotions?
No. Block’s recordings show expressive speech, but audible emotion does not demonstrate subjective feelings or consciousness. No independent test cited in the source establishes either.
Has anyone outside Block tested the model?
As of October 26, 2026, the source report says no outsider had run the model. Reporters assessed Block’s selected recordings, but there were no published independent benchmarks or public weights.
Can a human-sounding voice confirm who is calling?
No. A natural voice is not reliable proof of identity. For a request involving money or sensitive information, use a separately verified callback or another established check.
Is Dessa publicly available?
Block had not announced a public release date, weights or API in the source material. Access was described as a waitlist, with early access for a small group of AI safety researchers.
Source: ThorstenMeyerAI.com
