DeepSeek, the rising Chinese AI model, has been banned by the Australian Department of Home Affairs from using federal government devices. The Australian government claims that DeepSeek collects data and exposes it to “extrajudicial directions from a foreign government that conflict with Australian law”.
The announcement was made in a policy issued on February 4, and most government entities have been ordered to “identify and remove all existing instances of DeepSeek products, applications, and web services on all Australian Government systems and mobile devices” in addition to preventing further access to DeepSeek’s apps and services. Only agencies working with regulations or national security are allowed to access DeepSeek products. However, prior approval is required in such cases, and appropriate mitigations to protect sensitive data should be implemented.
Australia is following in the footsteps of several government bodies that are banning DeepSeek worldwide. Italian data watchdog GPDP had a more extreme reaction to the app and its shady data collection services, which resulted in a blanket ban on DeepSeek apps and services in the country. The GPDP had previously asked DeepSeek to disclose how they process and collect data but found the information provided by the company “completely insufficient”, leading to an immediate ban and an investigation.
Several US government agencies and the Taiwanese have also banned the service. India’s finance ministry has also issued a directive banning employees from using AI tools like DeepSeek and ChatGPT, as well as concerns regarding government documents and data confidentiality.
Despite its popularity, DeepSeek has been controversial, primarily due to China’s national security and counterespionage laws that force Chinese companies to give up user data if the government asks. Since DeepSeek stores user data in servers on Chinese soil, there’s little government or third-party organisations can do to prevent collected data from falling into Chinese hands.
Then, there are issues about shady practices from the model itself. The service has already suffered its first major data breach within a month of being launched to the public, and researchers have already started poking holes in its security measures. The model is also reportedly susceptible to jailbreak techniques like EvilBOT and Bad Likert Judge. Microsoft is also investigating allegations from OpenAI about DeepSeek misusing its API to extract data used by the former’s AI model.
In the News: Mistral Solutions hit by Bashe ransomware in a major cyberattack
