🔍 Read the full analysis: Block’s Dessa Speaks Naturally—but Voice Alone Is No Proof 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 that generates and listens to speech, and published demonstrations featuring laughter, interruptions and emotional cues. Those clips are company-selected: the model is not publicly available, and no outside researcher has tested its performance or safety. Its human-like voice also renews concerns about impersonation scams.
Block published demonstrations of its Dessa speech model on October 23, showing an AI that laughs, responds to emotion and speaks over a conversation, but the company has not released the model for independent testing. The demonstrations raise questions about whether more natural voice AI could make impersonation harder to detect, while the model’s performance and safeguards remain unverified outside Block.
Block describes Dessa as a model trained from scratch on human conversation audio, without text or pretrained models. The company says it can listen and speak at the same time, recognize cues such as sarcasm and laughter, and run faster than real time on a MacBook Pro M4 Max or better. Block has not published a latency figure or technical paper supporting those claims.
The recordings on Block’s project page include laughter during a sentence, a conversation about a forgotten film title and a response to an anniversary surprise. Block says the clips are live conversations and that nobody steers the model. They are still company-selected demonstrations, not independent evaluations. The model is not publicly available; Block has offered a waitlist and said a small group of AI safety researchers will receive early access.
Independent coverage has described both the model’s naturalness and its limitations. The Verge reported that some exchanges were awkward or uncanny, though Dessa sounded more natural than a comparison system in its assessment. TechCrunch said it was unclear whether emotional output might be guided by tags. As of October 26, Block had not disclosed the model’s size, benchmark results, license or public release date, according to CoinDesk.
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
A voice that handles interruptions and conversational cues could be more convincing in a phone call than a system that speaks in flat, carefully separated turns. That is a reasonable concern, not evidence that Dessa has been used in fraud. No public test has shown how well it can imitate a specific person, and Block has not said whether it can do so.
The broader risk exists independently of Dessa. The FBI’s 2025 Internet Crime Report, published in April 2026, counted 22,364 complaints mentioning AI and reported $893 million in losses associated with those complaints. The report also said victims lost more than $5 million to distress scams using cloned voices of loved ones. Those figures describe reported complaints and losses, not an estimate of Dessa-related activity.
For businesses and families, the immediate implication is practical: a familiar-sounding voice should not be treated as proof of identity when money, credentials or urgent action are involved. Call back using a number already on file, confirm sensitive requests through a separate channel and agree on a family code word. Realistic speech can make a conversation feel trustworthy; it cannot establish who is speaking.
Products Worth Considering
Sonos Era 100 SL Compact Microphone-Free Speaker, Black
Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
As an affiliate, we earn on qualifying purchases.
What Block Says Dessa Can Do
Block, the company behind Square and Cash App, announced the project on October 23. Co-founder Jack Dorsey described speaking with it as feeling like talking with a person, citing emotion, pauses and natural cadence. Dessa lead Ragavan Thurairatnam called the work “a GPT3 moment for speech” and said it could run on consumer GPUs. These are company leaders’ characterizations, not independently verified measurements.
Block says the model was trained using one NVIDIA GB200 NVL72 and hundreds of times fewer words than large language models require. It also says Dessa chooses its own voice and personality. The company has not given enough public technical information to evaluate those claims, establish what “hundreds of times fewer” is being compared with, or determine how the system behaves across languages and users.
The project sits at the intersection of speech technology and fraud prevention. Block argues that voice systems often miss emotional signals, such as a caller being in pain. Critics have raised the opposite concern: making artificial voices more human may lower listeners’ guard. The disagreement is about the uses and consequences of the same capability; neither side’s broader predictions have been established by the demonstrations.
“Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.”
— Jack Dorsey, Block co-founder
Independent Evidence Is Still Missing
The central limitation is that outside researchers have not publicly tested Dessa. There is no published paper, model card, benchmark suite or public model access identified in the source material as of October 26. Without those, outsiders cannot verify Block’s claims about training, speed, conversational ability or how the model compares with other systems.
Several safety and deployment details are also unanswered. Block has not specified the languages Dessa supports, whether it can imitate a particular person, or whether it uses watermarking or consent checks. The company says the model selects its own voice and personality, but that does not settle whether users could direct it to sound like someone else. The status and scope of the promised researcher access are not clear from the available information.
Nor do the demonstrations establish that the system understands emotion, experiences feelings or can reliably identify a caller’s emotional state. They show how selected outputs sound. The difference matters both for claims about the technology and for assessing how it might be misused.
Watch for Researcher Access
The next useful evidence would be independent researchers’ findings, a technical paper and a clear account of access, safeguards and evaluation methods. Block has said a small group of AI safety researchers will receive early access, but it has not announced a public release date. Until researchers can test the system and report results, claims about its general performance should remain attributed to the company.
For organizations handling sensitive calls, the model’s availability is not a prerequisite for strengthening identity checks. A voice alone is a weak basis for approving payments or changing account details; established callback procedures and separate verification channels offer a more dependable check. Families can likewise agree on a code word and verify urgent requests by calling a known number.
Block’s announcement also comes as rules for synthetic media develop. The source article says Article 50 of the EU AI Act applies from August 2, 2026, but it does not provide legal analysis of how that provision applies to Dessa or particular uses. The company’s own disclosures and future independent assessments will clarify what the model does and what protections accompany it.
Key Questions
What is Block’s Dessa?
Dessa is a speech-to-speech research model that Block says was trained from scratch on human conversation audio. The company says it can listen and speak at the same time and respond to conversational and emotional cues.
Can the public use or test Dessa?
Not based on the information available as of October 26, 2026. Block has not announced a public release and has described a waitlist and early access for a small group of AI safety researchers.
Has anyone independently verified the demonstrations?
The available clips were published by Block, and outside researchers have not publicly tested the model. Media outlets have listened to the demonstrations, but that is different from inspecting the model or independently benchmarking it.
Does Dessa prove that voice calls can no longer be trusted?
No. The demos do not establish that Dessa can impersonate a specific person or that it has been used in scams. They do reinforce a broader precaution: a familiar voice is not proof of identity, particularly for urgent requests involving money or account access.
What should people do about urgent requests by phone?
Hang up and call back using a number already known to be genuine. For family emergencies, a pre-agreed code word can help verify a caller; for financial or account changes, confirm through a separate, trusted channel.
Source: ThorstenMeyerAI.com
