A regulatory framework for AI models is being discussed in Australia. Google is also getting involved. According to a media report, the company has brought a copyright change into play: According to this, all available content should be able to be used for training AI models. If someone does not want to release their works for this, there should be an opt-out procedure according to Google.
According to the Guardian’s report, Google has called on Australian policymakers to “promote a copyright law that allows fair and fair use of copyrighted content to enable the training of AI models in Australia on a wide and diverse range of data, and at the same time to support working opt-outs for companies that prefer their data not to flow into AI systems.” When asked about this, Google is said to have referred to an older blog post. It emphasizes that publishers should have choice and control over their content, and with it the opportunity to participate in the value creation on the Internet.
New AI search complicates monetization
This reasoning is similar to that which Google also brought out in the dispute over the use of content in search results. The Australian government’s plans for a media law, which provided for the participation of publishers whose content Google uses in previews and as text excerpts, led to a demonstration of power by the group. Google argued, as did Meta for its own services, that their platforms first led users to the publishers’ websites and that they would already benefit from this. License payments were agreed upon, but the amounts are unknown.
In the blog post, Google also refers to the possibility of an opt-out, which is already being developed. Robots.txt works for Google search crawlers. Now it should also work for the crawlers of the providers of AI models. OpenAI has already announced how the web crawler can be blocked in a website’s robots.txt file.
In Australia, however, there are said to be talks between publishers and AI companies about payment if content is used to train the models. In addition to the training, the question is of course where large language models get the current information they need to answer questions. For example, if Google answers in the form of continuous text in its upcoming “Search Generative Experience”, then the links and thus the clicks for the publishers are missing, which is the only way to monetize their content.
Google also recently changed its terms of service, securing such extensive rights that you’d think the internet belongs to Google. It states: “For example, we collect data that is available online or in other public sources to train Google’s AI models and to further develop products and features such as Google Translate, Bard and Cloud AI. When your company information appears on a website , we can index them and display them in Google services.”
Go to home page
#Australia #Google #considers #copyright #law #optout #viable