A hot potato: Training the next generation of generative AI models requires a significant amount of high-quality, meaningful data that computers alone cannot generate. Now Mozilla wants to know if Microsoft intends to use user-generated information for algorithm coaching.
If generative AI technology is left to its own devices, there is a risk that it will increasingly lose value due to the feedback loop effect. Synthetic data may offer a potential solution to this problem, but in order to achieve the best training results, a significant amount of user-generated content (UGC) will likely be required in the near future. Microsoft has a large user base and could well choose to leverage this UGC to give itself a competitive edge over other big tech companies.
The Redmond company recently warned its users of the forthcoming changes in the Microsoft Services Agreement. This agreement governs how people use Microsoft’s online consumer products and services. The company provided a Summary of the main changes to the agreement, explaining privacy, content, code of conduct, Bing and of course AI services.
The updated document includes an entirely new section titled “AI Services” in which Microsoft describes the actions that users are allowed and prohibited to take with these services. The revised service agreement also states that Microsoft will “process and store your inputs to and outputs from the service” in order to monitor and prevent “abusive or harmful uses” of the service. In addition, privacy and Your Content sections have been expanded to include new AI-based services such as Bing Chat.
While the new Microsoft Services Agreement can be lengthy and tedious to read, it is expected to provide sufficient clarification regarding the company’s intentions with user data for AI training purposes. According to Mozilla, however, it still seems difficult to understand Microsoft’s true intentions, even after experts were brought in to review the new document.
In recent days, the open-source foundation has humorously urged Microsoft to be more transparent about its AI training practices. The company used its social media presence, particularly on X, to announce that its team of nine privacy experts were in attendance unable to determine if Microsoft intends to use personal data for AI training. These experts spoke out confusion and further asked clarification regarding the company’s new service contract.
Mozilla’s X campaign launched an official petition addressed Microsoft and asked a direct question: “Are you using our personal data to train AI?” The petition highlights that if four lawyers, three privacy experts and two activists were not able to demonstrate Microsoft’s intentions to see through the new agreement what chance the average citizen would then have?
While there are no guarantees, hopefully Microsoft can consider the petition if a sufficient number of users sign it. Meanwhile, Mozilla is actively promoting a more “responsible” and socially responsible approach to generative AI technology.