Dutch gov't wants to ban officials from using AI software
The Dutch government wants to ban government officials from using AI software like chatbots ChatGPT and Brand or image makers Dall-E and Midjourney for the time being. The risks in terms of privacy and copyright infringements are currently too high, State Secretary Alexandra van Huffelen (Digital Affairs) said in a draft proposal in the Volkskrant’s possession.
Van Huffelen drew up the proposal based on research by the office of State Attorney Pels Rijken and the Dutch Data Protection Authority (AP). They concluded that “non-contracted generative AI applications generally do not demonstrably comply with Dutch privacy and copyright legislation.”
According to Van Huffelen, it is “unclear” whether providers like Google or OpenAi sufficiently consider authors' rights. The language models behind chatbots like ChatGPT and Bard were trained using large amounts of texts, including books and newspaper archives on which copyright exists. The same applies to image AI. Many providers are not transparent about which works they use, Van Huffelen wrote.
She is also concerned about privacy. “A generative AI application can also derive very sensitive information from the interaction with the user.” And the answers provided by the AI program “could be used to make a decision about an individual,” which is another concern.
Van Huffelen added that she doesn’t want to write off generative AI use within the government entirely. She plans various experiments to see how government services can use the technology safely. The pilots should be ready by mid-2024, after which the government can draw up guidelines for the responsible use of AI. There will also be a training program for civil servants.
The Council of Ministers will soon discuss the draft proposal, according to the Volkskrant. If approved, the ban will apply to government services and their suppliers.