🔍 Read the full analysis: Can AI Sound Human Without Being Human? Block’s Dessa Makes You Ask on ThorstenMeyerAI.com
Get the latest gadgets 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 announced Dessa, a speech-to-speech research model, on October 23, 2026, and shared recordings that it says show natural, expressive conversation. No independent researchers have publicly tested the model, and Block has not released a paper, benchmarks, weights or an API. Its reported capabilities raise questions about both conversational AI and voice impersonation risks.
Block published demonstrations of Dessa, a research model designed to listen and speak at the same time, on October 23, 2026. The company says it trained the model from scratch on human conversation alone and that it can handle pauses, interruptions and vocal emotion; those claims have not been independently tested, and the model is not publicly available.
Block’s project page presents recordings in which Dessa laughs mid-sentence, responds to a surprise and helps a speaker recall a film title. The company says the clips are live conversations and that nobody steered the model during them. They are company-selected demonstrations, not a public evaluation or a substitute for outside testing.
Block describes Dessa as a speech-only, full-duplex model: it listens while it speaks, rather than relying on a conventional pipeline that first converts speech to text. The company says it was trained without text or pretrained models, using one NVIDIA GB200 NVL72, and can run faster than real time on a MacBook Pro M4 Max or better. Block has not provided a specific latency figure.
Several parts of the system remain undocumented publicly. As of October 26, Block had released no technical paper, model card, benchmark results, weights or API, and it had not announced a general release date. Access is limited to a waitlist and a small group of AI safety researchers, according to the source material. The Verge described some exchanges as “a little awkward and uncanny,” while reporting that the model sounded more natural than another system in one comparison.
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 Trust Risks
Dessa’s demos focus on conversational qualities—laughter, backchanneling and interruptions—that can make a voice interaction feel less scripted. Those features could make voice assistants easier to talk with. They also matter to fraud prevention because callers often use conversational cues to judge whether a voice sounds familiar or authentic.
The FBI’s 2025 Internet Crime Report, published in April 2026, counted 22,364 complaints mentioning AI, with $893 million in reported losses. The report also recorded more than $5 million in losses from distress scams involving cloned voices of loved ones. These figures describe complaints and losses reported to the FBI; they do not establish that Dessa was involved, or that this model has been used in fraud.
The practical concern is broader than one unreleased model: a convincing voice should not be treated as proof of identity. Organizations handling payment requests can use callbacks to verified numbers and written approval processes. Families can agree on a code word and independently call a known number if someone phones with an urgent request for money. These steps address existing voice-cloning risks, not a confirmed misuse of Dessa.
Products Worth Considering
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
Bose S1 Pro+ Portable PA Speaker
As an affiliate, we earn on qualifying purchases.
A Research Demo, Not a Release
Block, the company behind Square and Cash App, announced Dessa through a project page on October 23, 2026. Its description says the model learns from recorded human conversation and can interpret vocal expression directly. Block argues that current voice systems can miss emotion, such as a person speaking while in pain. That is the company’s rationale for the research; it does not establish how well Dessa handles such situations in practice.
Jack Dorsey, Block’s co-founder, wrote that talking with the model feels like talking with a person, citing its emotion, nuance, pauses and cadence. Dessa lead Ragavan Thurairatnam called it “a GPT-3 moment for speech” and said it could run on consumer GPUs. Those are statements from the project’s leadership, not independently verified performance findings.
Outside coverage has raised questions about the evidence behind the demos. TechCrunch reported that it was unclear whether emotional delivery might be guided by tags. CoinDesk noted the absence of model size, benchmarks and a public release date. The available material does not include an independent replication or a published account of the training data and evaluation methods.
“Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.”
— Jack Dorsey, Block co-founder
Independent Tests Are Still Missing
The central uncertainty is how Dessa performs outside Block’s recordings. There are no public benchmarks or independent test results in the material available as of October 26, 2026. The model’s size, training-data details, language coverage and release timetable also have not been disclosed there.
Block says Dessa chooses its own voice and personality, but it has not publicly clarified whether the model can imitate a specific person, what consent safeguards it uses, or whether its speech is marked or watermarked. The source material also does not establish whether vocal emotion is generated entirely by the model or partly shaped by other controls. No evidence provided links Dessa to an actual scam.
Outside Access Will Shape Evaluation
The next meaningful step would be independent access and documented testing. Researchers would need to assess conversational performance, latency, language support, data handling and safeguards, rather than relying only on launch recordings. Block has said a small group of AI safety researchers has early access, but it has not named a general release date or published a testing plan.
Until more information is available, readers and businesses should treat Dessa’s capabilities as claims illustrated by demos, not as established results. Existing safeguards for voice-based requests remain relevant regardless of whether Block releases the model more widely.
Key Questions
What is Block’s Dessa?
Dessa is a speech-to-speech research model that Block says can listen and speak at the same time and interpret vocal emotion. It is not publicly released.
Can the public use or test Dessa?
Not generally, based on information available on October 26, 2026. Block has described a waitlist and early access for a small group of AI safety researchers, but has not announced a public API or model weights.
Has Dessa been independently verified?
No independent evaluation is reported in the source material. The public evidence consists of Block’s demonstrations and media accounts of listening to them; no paper or benchmark results have been published there.
Does the announcement show that Dessa has been used for fraud?
No. The material describes general risks from voice cloning and reports FBI figures on AI-related complaints, but it does not link Dessa to any fraud case.
What should people do about urgent voice requests for money?
Do not rely on a familiar-sounding voice alone. Hang up and call a number you already know, and consider agreeing on a family code word for emergencies.
Source: ThorstenMeyerAI.com
