🔍 Read the full analysis: AI Voice And The Illusion Of Proof: Block’s Dessa Laughs And Hesitates 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 in which it laughs, pauses and responds during conversation. The demos are company-selected; no public model, technical paper or independent test is available, so its performance and safeguards remain unverified.
Block has published recordings of Dessa, a speech-to-speech research model that the company says can laugh, respond to emotion and talk over live conversation, raising questions about both more natural voice interfaces and impersonation scams. The recordings are Block-selected demonstrations, however, and the model has not been made publicly available for independent testing.
Block introduced the project on October 23. The company says Dessa was trained from scratch on recorded human conversation, without text or pretrained models, and can listen while speaking. Its demonstrations include a laugh during a sentence, a response to an anniversary surprise, and help recalling a film title. Those are examples supplied by Block, not independently verified results.
Block also says inference runs faster than real time on a MacBook Pro M4 Max or better, and that training used one NVIDIA GB200 NVL72. It has not provided a latency figure, model size, benchmark results, technical paper or model card. Block has announced no public weights or API; access is described as a waitlist and early access for a small group of AI safety researchers.
Outside coverage has noted both the natural-sounding exchanges and their limits. The Verge reported that some conversations were awkward and uncanny, while TechCrunch said it was unclear whether emotion in the demos might be steered by tags. CoinDesk reported that Block had not given a public release date. These observations concern the recordings, not tests of the underlying model.
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
More fluid conversation could make voice systems easier to use, particularly when they need to handle interruptions, uncertainty or emotional cues. Block says the ability to notice those cues could help a voice AI respond appropriately when someone is in pain. But the same characteristics could make a synthetic voice harder for a listener to distinguish from a person.
The risk is not hypothetical in the broader category of voice fraud, though the supplied data does not show that Dessa has been used in scams. The source cites the FBI’s 2025 Internet Crime Report, published in April 2026, as recording 22,364 complaints mentioning AI and $893 million in reported losses. It also says victims lost more than $5 million to distress scams involving cloned voices of loved ones. Those figures describe reported incidents, not a measure of Dessa’s capabilities or its effect on fraud.
For companies, the demonstrations are a reminder that a familiar or convincing voice should not be treated as proof of identity. Payment requests and sensitive account changes need verification through a separate, established channel. Families can agree on a private code word and call back using a number already known to them rather than one supplied during an urgent call.
Products Worth Considering
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
Sonos Era 100 SL Compact Microphone-Free Speaker, Black
As an affiliate, we earn on qualifying purchases.
Block’s Claims and the Evidence
Block is the company behind Square and Cash App. Its project page describes Dessa as the work of a small research lab associated with the team from Toronto-based deep-learning startup Dessa, which Square acquired in 2020, according to CoinDesk. The shared name refers to the research project in Block’s announcement, not evidence that the old startup’s products or technology are being released here.
Block co-founder Jack Dorsey wrote that talking with Dessa feels like talking with a person, citing emotion, nuance, pauses and cadence. Team lead Ragavan Thurairatnam called the work “a GPT3 moment for speech” and said it could run on consumer GPUs. These are company and team characterizations, not independent findings. The demos offer a limited view of interactions selected and published by Block.
Several basic questions remain open in the available material: Block has not listed the model’s languages, published a license, or said when a broader release might happen. The company says Dessa picks its own voice and personality, but the source material does not establish how it handles requests to imitate a specific person or what safeguards it uses.
“Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.”
— Jack Dorsey, Block co-founder
Testing, Safeguards and Release
No outside researcher or journalist has publicly tested the model itself in the source material available as of October 26. Without access to the system, recordings cannot establish how consistently it handles interruptions, whether the exchanges were shaped by prompts or other controls, or how it performs outside the showcased situations.
Block has not published a technical paper, model card, benchmark results, license or release date. It is also unclear which languages Dessa supports, whether its output is watermarked, what consent checks apply, and whether it can imitate a particular person. The source material does not establish whether any protections against impersonation are present.
The reported FBI figures provide a wider picture of AI-related complaints and losses, but they do not show that Dessa has been used for fraud or quantify how much conversational realism changes scam success. Claims about that specific effect remain untested.
Independent Access Is the Next Test
The next meaningful milestone would be access for independent researchers, alongside documentation describing training, capabilities and safeguards. Block has said a small group of AI safety researchers has early access, but has not announced when findings, a wider release or a public API might follow.
Until then, readers should treat the recordings as demonstrations rather than proof of broad performance. Organizations handling payments or account access can reduce exposure by requiring verification through known, separate channels. People receiving urgent requests for money should hang up and call the person back on a number they already trust; a convincing voice alone cannot confirm who is calling.
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. The company has shared recordings of conversational exchanges, but has not made the model generally available.
Has Dessa been independently tested?
Not in the source material available as of October 26, 2026. Reporters have listened to Block’s recordings, but no outside party is reported to have tested the model directly.
Does the announcement prove that Dessa can clone a specific person?
No. The available information does not establish whether it can imitate a specific person or what consent and identity safeguards it uses. Block says the model picks its own voice and personality.
How should people respond to a suspicious voice call?
Do not treat a familiar voice as proof of identity. For urgent requests involving money, hang up and call back using a number already known to you; families may also agree on a private code word.
Source: ThorstenMeyerAI.com
