🔍 Read the full analysis: Block’s Dessa Laughs And Hesitates Like A Person. Does That Mean Anything? 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 introduced Dessa, a speech-to-speech research model whose company-selected demos feature laughter, interruptions and conversational pauses. The model has not been independently tested or publicly released, so its capabilities and safeguards remain unverified.
Block has introduced Dessa, a speech-to-speech research model shown in company demos laughing, pausing and interrupting in conversation. The October 23 announcement has drawn interest because those behaviors may make voice AI feel more natural—and potentially make synthetic voices harder to distinguish—but no outside researcher has tested the model, and Block has not released it publicly.
Block says Dessa was trained from scratch using human conversation, without text or pretrained models. The company describes it as a full-duplex system, meaning it can listen while speaking, respond to interruptions and use conversational signals such as laughter or backchanneling. Block also says it runs faster than real time on a MacBook Pro M4 Max or better, but has not provided a specific latency measurement.
The evidence available so far consists of recordings selected and published by Block. In them, the model laughs during a sentence, helps retrieve a forgotten film title and responds to an anniversary surprise with what the source describes as teary-sounding gratitude. Block says the clips are live recordings and that nobody steered the model. Those demonstrations show what the company chose to present; they do not establish how consistently Dessa performs in other conversations.
Independent reporting has noted both its natural-sounding moments and its limits. The Verge described some exchanges as occasionally awkward and uncanny, while finding Dessa more natural-sounding than another model in a comparison. TechCrunch said it was unclear whether emotional expression might be influenced by tags. As of October 26, Block had published no technical paper, model card, benchmark results, weights or API, and had announced no public release date. Access is described as a waitlist and early access for a small group of AI safety researchers.
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
Humanlike Voices Raise Fraud Risks
Dessa’s conversational style matters because voice realism has both useful and risky applications. Natural pauses, laughter and responses to interruption can make an automated support call feel less mechanical. The same qualities could make an impersonation more convincing if a person believes they are speaking with a relative, colleague or company representative. That risk does not depend on Dessa being available: voice-cloning fraud already exists.
The FBI’s 2025 Internet Crime Report, published in April 2026, recorded 22,364 complaints mentioning AI and $893 million in reported losses. The FBI also said victims lost more than $5 million to distress scams in which cloned voices mimic a loved one in trouble. Those figures cover reported complaints and losses, not every incident, and they do not show that Dessa was used in any fraud. They provide context for why a more natural-sounding voice model is receiving scrutiny.
For readers, the practical point is not that every humanlike voice is fraudulent. It is that a familiar-sounding voice should not be treated as proof of identity. A call demanding urgent payment, secrecy or personal information warrants verification through a known number or a separate channel. Families can agree on a code word for emergencies; businesses can require independent confirmation for payment changes and transfers.
Products Worth Considering
Sonos Era 100 SL Compact Microphone-Free Speaker, Black
Bose SoundLink Flex Bluetooth Speaker (2nd Gen), Portable Waterproof Speaker, Black
Sony WH-1000XM4C Wireless Noise-Cancelling Headphones, Black
As an affiliate, we earn on qualifying purchases.
What Block Has—and Has Not—Shown
Block, the company behind Square and Cash App, posted a project page for Dessa on October 23, 2026. Co-founder Jack Dorsey described talking with it as like talking with a person, citing emotion, nuance, pauses and natural cadence. Dessa lead Ragavan Thurairatnam called the work a “GPT3 moment for speech,” according to the supplied source material. Those are company and team characterizations, not independent findings about the model’s performance.
Block says Dessa learned from hundreds of times fewer words than large language models require and was trained using a single NVIDIA GB200 NVL72. The company has not provided enough technical detail for outside observers to verify those claims or compare the system with other models. Its stated goal includes exploring how efficiently speech models can learn new languages; the languages supported have not been listed.
The announcement has prompted sharply different reactions. Some listeners have praised how human the demos sound; others have raised concerns that realism could lower a caller’s guard. Block has framed emotional awareness as a potential benefit, arguing that a voice system should recognize cues such as pain or frustration. The same capability can support more responsive interactions while also increasing the importance of consent, disclosure and identity checks.
“Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.”
— Block co-founder Jack Dorsey
Independent Testing Is Still Missing
Dessa’s capabilities remain unverified outside Block. There is no published paper, model card, benchmark suite or independent evaluation in the supplied reporting, and no public model access for researchers to reproduce the demonstrations. The company has not disclosed the model’s size, a release schedule or a specific latency figure.
Important safety and consent details are also unanswered. The available information does not establish whether Dessa can imitate a particular person, what checks prevent that, whether its output is watermarked, or what consent is required for voices represented in training or generated by the system. Block says the model chooses its own voice and personality, but that statement does not resolve how it handles requests to sound like an identifiable individual.
Nor is it clear how reliably the model handles different languages, accents, noisy calls or high-pressure conversations. The company’s demos are a limited sample, and reports of awkward exchanges suggest performance may vary. No evidence in the source material links Dessa to a fraud incident; the concern is about potential misuse of realistic voice technology generally.
Release Details and Safeguards Ahead
The next meaningful test will come if Block gives outside researchers broader access or publishes technical documentation. Independent evaluations could examine conversational performance, latency, language coverage, consent controls and whether listeners can identify generated speech. Until then, claims about the model’s abilities should be understood as Block’s description of a research system, supported by selected demonstrations rather than public testing.
Block has not announced a public release date, so it is unclear when businesses or consumers might be able to use Dessa. Its access page points to a waitlist and limited early access for AI safety researchers. Any future launch details, technical disclosures and safeguards would help clarify whether the demonstrations translate into dependable use beyond curated clips.
In the meantime, organizations handling payments or sensitive requests can use procedures that do not rely on a caller’s voice alone: call back using a number already on file, confirm unusual instructions through a second channel, and disclose when a caller is interacting with AI where required. Individuals can take similar steps by hanging up and calling a known number when a request creates urgency. These precautions address existing voice-fraud risks, whether or not Dessa becomes publicly available.
Key Questions
What is Block’s Dessa?
Dessa is a speech-to-speech research model that Block says was trained from scratch on recorded human conversation. The company says it can listen and speak at the same time and produce conversational cues such as laughter and pauses.
Can the public use or independently test Dessa?
Not based on information available as of October 26, 2026. Block has not announced public model weights or an API, and the reported access is limited to a waitlist and a small group of AI safety researchers.
Does Dessa prove that AI can feel emotions?
No. A voice that sounds emotional does not establish that a model experiences emotion. The published clips show expressive speech; they do not demonstrate feelings or humanlike understanding.
Has Dessa been used in a reported scam?
The supplied reporting does not identify a fraud case involving Dessa. The concern is broader: voice cloning is already used in some scams, and more natural conversational behavior could make synthetic voices harder for people to assess.
How can people verify an urgent call about money or an emergency?
Hang up and call the person or organization using a number you already know, rather than one provided during the call. Families may also agree on a private code word; businesses can confirm payment instructions through a separate channel.
Source: ThorstenMeyerAI.com
