Documents reveal that staff members at the Department of Defence accessed the servers of artificial intelligence (AI) chatbot, ChatGPT, without proper approval. The department has since taken measures to restrict access to the web domain of OpenAI, the owner of ChatGPT, in order to prevent any potential data breaches or privacy concerns.

According to information obtained through a freedom of information request, Department of Defence devices, including computers and smartphones, connected to webpages with the OpenAI.com domain a total of 5630 times between December 1st, 2022, and June 30th, 2023. The OpenAI.com domain hosts various AI products, including ChatGPT, which was launched on November 30th, 2022.

It is worth noting that the connections made by Defence Staff were through the online web interface of OpenAI.com, specifically excluding connections made through third-party services or smartphone apps provided by OpenAI.

In a letter accompanying the released documents, the department’s accredited decision-maker, David Evans, confirmed that the department has not approved access to OpenAI’s products and has implemented restrictions to prevent the loss of control of classified or privacy information. The 5630 requests include instances of connections made prior to the decision to restrict access.

There is currently no government-wide advice for federal departments regarding the use of generative AI products like ChatGPT, although it has been suggested that a full evaluation of potential risks should be conducted. Greens senator David Shoebridge expressed concern over the lack of a credible government-wide policy to address security concerns and called for new policies and regulations to protect classified information.

Source: This article is reproduced from our sister publication Crikey.

APS issued official guidance on workplace use of ChatGPT, Bard AI and Bing AI.

Sources:
– Crikey

The post Department of Defence Staff Access AI Chatbot Servers Without Approval appeared first on TS2 SPACE.