In early July 2023, a few quiet words changed in the research and development section of Google's privacy policy, and most readers passed over them without a second look. The edit, first noticed by Gizmodo and carried forward by Android Authority, says publicly available information may help train Google's AI models and power products like Bard and Cloud AI. It reads as a small revision on paper, yet it touches an older human question, what becomes of what we leave in public view, now made pressing by machines that learn from it. Whether that reach is intended, and how far it will be honored,
Google's Updated Privacy Policy Says It Can Use Public Data to Train Its AI Models
Not just who can read what you write, but what it becomes
Mark
So what actually changed in Google's policy?
Mimi
A sentence in the research and development section. It now says Google uses publicly available information to help train its AI models, and it names Google Translate, Bard and Cloud AI capabilities.
Mark
And before?
Mimi
It said "language models," not "AI models," and it only named Google Translate. So the wording is wider and the product list is longer.
Luke
Worth being careful here. The text change is verifiable, since it's a comparison of two versions. But the reading that Google claims rights over the whole public web is an interpretation of the wording, not something Google has said.
Mark
Why does the "public" part matter so much?
Mimi
Policies like this usually describe data from the company's own services. This one reads as though it reaches anything on the open internet.
Mark
Is that really new, though? People know public posts are public.
Mimi
The article's point is that visibility and use are different things. Being readable by anyone is not the same as being material for a product.
Luke
And we don't know whether Google's practices changed at all. The piece doesn't say. It reports a policy edit, not a new data collection program, and it has no comment from Google.
Mark
What about the lawsuits?
Mimi
The article says tools like Bard, ChatGPT and Bing Chat rely on scraped web content, which can include others' intellectual property, and that suits accusing them of theft already exist.
Luke
Right, but those are accusations. Nothing in this piece says how any of them will be decided.
The Pulse
- A few quietly changed words in Google's privacy policy have set off alarm about where its AI training data really comes from.
- The new clause points to 'publicly available information' and no longer limits the scope to Google's own platforms, raising fears of web-wide harvesting.
- Critics are weighing the gap between what people expect when they post online and what the policy now lets Google turn that content into.
- Google has not yet responded, leaving users to parse the wording themselves while pending lawsuits over scraped content loom over the debate.
- The story is heading toward either a clarifying statement from Google or a legal reckoning that could define how such policy language is judged.
In early July 2023, a few quiet words changed in the research and development section of Google's privacy policy, and most readers passed over them without a second look. The edit, first noticed by Gizmodo and carried forward by Android Authority, says publicly available information may help train Google's AI models and power products like Bard and Cloud AI. It reads as a small revision on paper, yet it touches an older human question, what becomes of what we leave in public view, now made pressing by machines that learn from it. Whether that reach is intended, and how far it will be honored, remains for Google to say and for the courts to test.
Over a weekend in early July 2023, Google edited its privacy policy in a way almost no one noticed. The change sits in the section explaining how the company uses information for research and development. Gizmodo spotted it, and Android Authority followed with the story.
The revised text says Google uses publicly available information to help train its AI models and to build products such as Google Translate, Bard and Cloud AI capabilities. The earlier version was narrower. It mentioned "language models" instead of the broader "AI models," and it named only Google Translate. Bard and Cloud AI are new additions.
The concern is less about the words themselves than their reach. A privacy policy usually describes how a company handles data on its own services, but this clause reads as if Google may draw on anything across the public internet. That shifts the familiar idea that what you post can be seen by others into a new question: what that material might become once it is used to train an AI.
The article sets the change against a larger backdrop. AI tools such as Bard, ChatGPT and Bing Chat are described as gathering information from the web, sometimes including other people's intellectual property, and lawsuits already accuse such tools of theft.
The reporting stays with the policy text. It includes no comment from Google and does not explain how the clause is applied in practice. What comes next depends on whether Google clarifies its wording and how the pending lawsuits over scraped content are resolved.
Over a weekend in early July 2023, Google changed a few words in its privacy policy, and almost nobody noticed. The edit sits in the section on research and development, the part of the policy that explains how the company uses the information it collects. It was spotted by Gizmodo, and Android Authority then picked up the story.
The section now says Google uses information to improve its services and to develop new products, features and technologies that benefit its users and the public. Then comes the sentence that drew attention: as an example, Google says it uses publicly available information to help train its AI models and to build products and features such as Google Translate, Bard and Cloud AI capabilities.
The previous version was narrower in two ways. It referred to training "language models" rather than "AI models," a broader term. And it named only Google Translate. Bard, Google's chatbot, and Cloud AI capabilities are new to the list.
Android Authority describes the wording change as slight but concerning. The concern, as Gizmodo framed it and Android Authority repeated, is the scope of the clause. A privacy policy normally covers how a company handles data posted on its own services. This one reads as though Google is reserving the right to collect and use information from anywhere on the public internet, not just from its own platforms.
That raises a distinction the piece draws out. Most people already assume that anything they post online can be seen by others. The new language adds a question of use: not only who can read what you write, but what that writing might be turned into once it becomes training material for an AI product.
The article places the change in a wider context. Bard, ChatGPT, Bing Chat and similar AI tools that provide real-time information are described as working by scraping information from the internet. Some of what they draw on can be other people's intellectual property. Lawsuits are already pending that accuse these tools of theft, and the article expects more to follow.
The reporting is limited to the policy text itself and how it compares with the earlier version. It does not include a response from Google, and it does not say how or whether the company applies the clause beyond what the wording states. What to watch is whether Google clarifies the language, and how the ongoing lawsuits over AI training on scraped content play out, since those cases may shape how policy wording like this is judged.
Notable Quotes
Google uses publicly available information to help train its AI models and build products like Google Translate, Bard, and Cloud AI capabilities.— Google privacy policy, research and development section
Want the full story?
Read the original at Android Authority ↗
Technology