- Mullvad VPN has deployed an upgrade to its functionality which protects against the monitoring of AI
- The second version of the defense against traffic guided by AI (DAITA) is now available for all VPN applications
- Mullvad published Daita in May 2024 as “the first step” of the battle against the analysis of sophisticated trafficking
Mullvad VPN has just unveiled a considerable upgrading of its protections against AI monitoring.
The second version of his defense against traffic guided by AI (DIAT) is now available for all VPN applications, making two major improvements that “make attackers targeting Daita more difficult”.
One of the best VPN suppliers for users concerned for confidentiality, Mullvad published Daita in May 2024 as “the first step” of the battle against a sophisticated analysis of traffic propelled by new generative AI software.
Daita 2.0
Built entirely on an open source framework, DAITA modifies the appearance of data packets sent to the VPN network to prevent snoopers from retracing the activity to you.
He does this by doing all the data packages of the same size, adding random background traffic and deforming the models.
According to Jan Jonsson, CEO of Mullvad VPN, this capacity is critical, because AI can be used to analyze traffic models. In simpler terms, this means that your activities can be linked to you without the need to break the encryption offered by the virtual private network (VPN).
“The AI will be used both to collect data on people and to analyze this data. This is why we developed DAITA, to counter immediately at the time of data collection,” said Jonsson in October.
On March 28, 2025, however, Mullvad published the second version of Daita on all its platforms to reduce attacks against the functionality itself.
Daita version 2 now available on all platforms: https://t.co/b77nm95uhkMarch 28, 2025
According to Mullvad’s own words, “Daita version 2 brings two major improvements: a great reduction in air traffic and dynamic configurations that vary the characteristics of the VPN tunnel between connections.”
The first improvement refines one of the secret practices that Daita uses to confuse the monitoring of AI – the addition of dummy packages to distort network models.
Daita 2.0 now inserts half of the quantity of these packages but promises to provide the same level of defense. “As a customer using DAITA, the immediate advantage is an improved speed,” said the supplier.
In the new version, two VPN customers visiting the same web page will now produce various essential data flows to make attackers more difficult to get around Daita protections. A new configuration is also selected each time a device recreates its VPN connection.
If you want more technical details on the functioning of Daita, I suggest reading this post by cybersecurity researchers from the University of Karlstad in this area.
Mullvad’s commitment to protect your data from AI monitoring does not stop there either. The team is already working on a new version of Daita.
“Daita version 3 is already on the roadmap and will introduce a new type of defense alongside those existing. Look at this space for more updates while we make the state of the defense of accessible and efficient network traffic.”
