Artificial intelligence is taking another step toward human-like communication with the launch of Dessa, a new AI voice model developed by Block, the financial technology company led by Jack Dorsey. Unlike conventional voice assistants, Dessa is designed to learn and communicate directly through speech rather than relying on text-based processing.
The model has attracted attention for its remarkably natural conversations, emotional expressions and ability to respond in real time. However, its launch has also raised concerns about phone scams, AI-generated deception and the growing difficulty of distinguishing artificial voices from real people.
What Makes Dessa Different?
Most traditional AI voice systems follow a process in which spoken words are converted into text, processed by a language model and then transformed back into speech. Block says Dessa takes a different approach by learning directly from human conversation, without text-based training or pre-trained models.
According to the company, this approach allows the model to preserve important aspects of natural speech, including tone, timing, pauses, humour and emotional cues.
Dessa also supports full-duplex communication, meaning it can listen and speak simultaneously. This enables it to respond more naturally, interrupt appropriately and react during conversations instead of waiting for a speaker to finish every sentence.
Demonstrations released by the company show the model laughing, expressing apparent sadness and participating in interactive guessing games. These examples suggest that speech-based AI could become more conversational and engaging than many existing voice assistants.
Why Are Users Concerned?
Despite the technical achievement, Dessa has received a mixed response online. While some people are impressed by its capabilities, others worry that highly realistic AI voices could create serious problems for personal security and online trust.
One of the biggest concerns is the potential rise of sophisticated phone scams. Criminals already use AI tools to imitate voices and impersonate family members, colleagues or trusted organizations. A model capable of producing natural conversations could make fraudulent calls more convincing and difficult to identify.
People may receive calls that sound as though they are coming from someone they know. If an AI system can reproduce realistic emotional reactions, pauses and conversational patterns, victims could find it harder to recognize suspicious interactions.
Older adults and people who are less familiar with AI technology may be particularly vulnerable to such tactics. However, the actual impact of Dessa on fraud remains uncertain, and the model’s release does not by itself demonstrate that it has been used in scams.
Another concern is the potential for emotional dependence on AI companions. As voice assistants become more expressive and responsive, some users may develop stronger emotional connections with them. Critics worry that this could encourage social isolation or blur the boundaries between human relationships and artificial interactions.
Impressive Technology, but Important Questions Remain
Dessa’s development also raises questions about how AI systems should communicate their identity. If an artificial voice sounds convincingly human, should it be required to disclose that it is AI at the beginning of every conversation?
Clear disclosure, safeguards against impersonation and restrictions on malicious use could become increasingly important as speech-generation technology advances.
There are also questions about privacy, consent and the handling of conversational data. Developers will need to demonstrate how their systems recognize emotional cues, protect sensitive information and prevent their technology from being misused.
Block has presented Dessa partly as a potential step toward safer voice AI. The company argues that conventional systems can miss emotional signals when speech is reduced to text, potentially causing an assistant to respond inappropriately when someone is distressed or in pain.
The company is initially offering access to a small group of AI safety researchers rather than releasing the model broadly. This limited approach could help researchers examine its capabilities and identify risks before wider deployment.
What Comes Next for Dessa?
The launch of Dessa highlights a broader shift in artificial intelligence: moving beyond systems that simply generate spoken answers toward models designed to understand and reproduce the rhythm of human conversation.
For Block, the technology could eventually support more natural customer service, interactive assistants and other voice-driven applications. Its ability to run locally on capable consumer hardware could also create opportunities for applications that do not depend entirely on remote computing infrastructure.
However, technical progress alone will not determine whether the technology succeeds. Public confidence will depend on how effectively developers address fraud, privacy, transparency and responsible use.
Dessa remains an early research preview, and its long-term performance and safety are yet to be established independently. The concerns surrounding its launch should therefore be understood as potential risks rather than proof that the model has already caused widespread harm.
Conclusion
Jack Dorsey’s Block has introduced an ambitious approach to voice AI with Dessa, demonstrating how artificial intelligence can learn directly from speech and produce more natural, expressive conversations.
Yet the technology presents a difficult trade-off. The more convincingly AI can imitate human communication, the greater the need for safeguards that protect people from deception and manipulation.
Dessa could represent an important milestone in conversational AI, but its ultimate success will depend on more than how human it sounds. The real challenge is ensuring that more realistic AI communication also becomes safer, more transparent and worthy of public trust.