🔍 Read the full analysis: Block’s Dessa Shows How AI Can Sound Human Without Settling The Question on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Block announced Dessa, a research model designed to listen and speak at the same time, and shared recordings that include laughter, interruptions and emotional speech. The demos are company-selected and no independent testing, technical paper or public model access has been reported, leaving its capabilities and safeguards unverified.
Block has demonstrated Dessa, a research model designed to hold spoken conversations with human-like timing and vocal expression, but no independent evaluation or public release has been reported. The October 23 announcement has drawn attention both to the model’s conversational abilities and to the possibility that more natural-sounding AI could make voice impersonation harder to detect.
Block says Dessa was trained from scratch on human conversation alone, without text or pretrained models. The company describes it as a full-duplex system: it can listen while speaking, interrupt, or provide conversational backchannels. Its demonstrations include laughter during a sentence, a response to a surprise, and help recalling a film title. These are recordings selected by Block, not results from an independent test.
Block also says the model runs faster than real time on a MacBook Pro M4 Max or better and was trained using one NVIDIA GB200 NVL72. The company has not supplied a latency figure, model size or benchmark results. It has announced no public API or model weights; access is described as a waitlist and early access for a small group of AI safety researchers.
Coverage of the demonstrations has been mixed. The Verge reported that some exchanges could feel awkward or uncanny, while finding Dessa more natural than another model featured in a comparison. TechCrunch said it was unclear whether emotion in the samples was directed by tags. As of October 26, no technical paper, model card, license or public release date had been identified.
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
Natural Speech Raises Fraud Risks
The central issue is not whether a model sounds emotional in a recording; it is whether listeners can reliably tell when a voice is synthetic. A more fluid system that handles interruptions and conversational cues could make a call feel more ordinary, potentially reducing one signal people use to judge whether a caller is genuine. That is a risk raised by the demonstrations, not evidence that Dessa has been used in fraud.
Voice impersonation scams predate this announcement. The FBI’s 2025 Internet Crime Report, published in April 2026, counted 22,364 complaints mentioning AI and $893 million in reported losses. The report also recorded more than $5 million in losses to distress scams using cloned voices of loved ones. Those figures describe reported complaints and losses across AI-related incidents; they do not measure Dessa’s impact or establish that every case involved voice cloning.
For organizations, the development is a reminder that a familiar-sounding voice should not, on its own, authorize a payment or account change. Independent verification through a known contact channel can reduce reliance on voice familiarity. Families can likewise agree on a code word and call back using a number they already trust. These are practical precautions against voice impersonation generally, not responses to confirmed misuse of this model.
Products Worth Considering
Jackery Explorer 300 Portable Power Station, 292Wh
Jackery HomePower 3600 Plus Portable Power Station, 3584Wh, 3600W
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
As an affiliate, we earn on qualifying purchases.
What Block Has Demonstrated
Block, the company behind Square and Cash App, published a project page for Dessa on October 23, 2026. The project is associated with a small research lab founded by former members of Toronto-based deep-learning startup Dessa, which Square acquired in 2020, according to CoinDesk. Block’s co-founder Jack Dorsey described the experience as “full of emotion, nuance, pauses, and a natural cadence.”
The company’s stated approach differs from common speech systems that convert speech to text before generating a response. Block argues that this kind of pipeline can discard vocal information such as emotion, sarcasm and laughter. Dessa is presented as hearing and producing those qualities directly. The claim is that this design supports more natural interaction; the public evidence remains limited to demos.
Block’s lead on Dessa, Ragavan Thurairatnam, called it “a GPT3 moment for speech” and said it could run on consumer GPUs. That is an attributed assessment, not an independently established comparison. The company has also identified learning new languages efficiently as an open question and says Dessa chooses its own voice and personality.
“Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.”
— Jack Dorsey, Block co-founder
Performance and Safeguards Remain Unknown
No outsider has been reported to have run or independently tested Dessa. That leaves key questions about reliability, latency and conversational performance unanswered, including how it behaves across accents, languages and less curated conversations. Block has not published the materials normally used to assess a model’s technical design and limitations, such as a paper or model card.
It is also unclear whether the model can imitate a specific person, what consent checks it uses, or whether its output includes watermarking or another disclosure signal. Block says Dessa selects its own voice and personality, but that does not resolve those questions. There is no reported evidence that the system has been used in scams or that it can reproduce a particular person’s voice.
The demonstrations do not establish that the model is emotionally aware, broadly intelligent, or suitable for unsupervised use. Nor do they establish that every exchange is fully spontaneous. Reporters have raised questions about whether some emotional qualities may be guided by tags; Block’s public materials, as described in the source reporting, do not settle that point.
Independent Access Is the Next Test
The next meaningful evidence would be access for independent researchers, publication of technical documentation and testing beyond Block-selected recordings. Those steps could clarify how the model handles interruptions, emotional cues, languages and attempts to prompt particular voices. No public release date has been announced, so the timing of that assessment is unknown.
Until more information is available, readers should treat claims about Dessa’s speed, training efficiency and naturalness as Block’s claims. Companies and families can review their existing protections against impersonation now: verify consequential requests through a previously known number or a separate channel, and avoid making decisions under pressure based only on a caller’s voice. The announcement makes the questions more immediate, but it does not establish that Dessa itself has reached general users.
Key Questions
What is Block’s Dessa?
Dessa is a speech-to-speech research model that Block says can listen and talk at the same time and produce conversational cues such as laughter and pauses.
Can the public use Dessa?
Not based on the information available as of October 26, 2026. Block has described a waitlist and early access for a small group of AI safety researchers, but has not announced public weights, an API or a release date.
Has Dessa been independently tested?
No independent test is reported in the source material. The public demonstrations are recordings selected by Block, and no paper, model card or benchmark results have been identified.
Does Dessa prove that AI can feel emotion?
No. Expressive speech in a demonstration does not establish that a system experiences emotion or has human-like understanding. The model’s internal capabilities remain unverified by public testing.
What can people do about voice impersonation scams?
Do not rely on a familiar voice alone for urgent requests involving money or sensitive information. Call back using a number you already know, verify through another channel, and consider agreeing on a family code word.
Source: ThorstenMeyerAI.com
