Large-scale language models and machine learning are new inventions by historical standards, but are already familiar to everyday cultures. Collectively referred to as “AI,” these technologies have launched Pandora's box of privacy concerns.
If AI is old, social media is ancient history. However, the privacy risks associated with social media use have grown since the first latest social media platforms came onto the scene about 20 years ago. These risks have increased gradually in recent years as so-called AI models began to be integrated and trained on social media platforms.
Researchers at Incogni have prepared a privacy ranking for the new social media platform for 2025 and expanded the standard to include concerns about LLM and so-called generation training. This year's rankings stand out by taking a more nuanced approach than previous research, expanding their scope to the end-user experience in collecting and analyzing privacy-related information.
The study looked at the top 15 social media platforms in monthly user counts and ranked them according to 14 criteria across six categories.
- AI integration and training
- Violation of privacy-related regulations
- Data collection
- User controls and consent
- Transparency
- User-Friendly
The results were appropriately weighted to generate an overall privacy ranking and ordered the platform according to the extent that poses privacy risks to users.
As expected, Meta's products (Facebook, WhatsApp, Instagram, and Facebook Messenger) and Bytedance's Tiktok close at the bottom of the rankings. But surprisingly, less popular platforms like Discord, Pinterest, and Quora have done relatively well.
Having an overall ranking like this is great, but it is a detailed analysis and has proven to be the most useful in making informed decisions about which platforms are reliable. For example, Pinterest is positioned so highly as a whole, it may not be the best option for users who are particularly concerned about collecting and sharing data. This is because Pinterest is at its worst.
The correlation between the overall location of the platform and its performance in the “AI and Personal Data” category is generally much stronger. This category covers criteria regarding whether the platform reserves the right to train its own entities or other entities in user data, and whether it provides users with a mechanism to opt out of such data use.
Apart from introducing criteria for AI-related categories, researchers added subjective aspects to their analysis. Subjective, yet quantified (for example, via the Dale Charle Formula): “User kindness and accessibility” criteria captures how difficult it is for users to parse relevant privacy policy documents, and how many individual steps a user has to take to delete an account.
This is an important part of the analysis as it helps to ground your research from the typical user's perspective. Concepts such as “user-friendly” and “accessibility” depend heavily on the extent to which actual users can reasonably expect to collect and process the information they need to make informed decisions. Although your privacy policy may contain all the right information, if inexplicably to a university-educated reader, it will not be able to perform core functions and will convey notable details to the average user.
Incogni head Darius Belejevas said this.
With all the spread of your imagination, there is no mainstream social media platform that allows you to respect privacy. However, there are social media platforms that respect the privacy of their users. Like projects like Mastodon, Pixelfed, and Nostrix and Matrix, the ActivityPub protocol that allows them to consolidate is a great example. However, all of these platforms share a common challenge. It is a low intake among everyday users. That is, network effects.
continuation:
The reality is that people want the connections and distractions that mainstream social media platforms promise. Protecting privacy is a desirable selling point, and being one of the things we all can do to shake up the market for a more user-friendly future. The first step to doing that is to understand the privacy risks associated with your current platform.
This study brings to the forefront a phenomenon that affects many aspects of this sometimes vague concept of “privacy.” Its increase and accelerated complications. Like newer than the personal industrial exploitation boom of the early 2000s, now, looking back, it clearly appears to be a simple time. Well beyond the harvest and analysis of personal data entered and generated through interactions with these platforms, these companies are currently using things they can find and guess to follow users around the web, survey devices, and train a variety of “AI” models.
All of this represents a double opening from the privacy risk landscape. On the one hand, the streams of personal data that enter these social media platforms have increased throughput as numbers increase. Meanwhile, streams of RAW, processing and inferred personal data that leave the same platform are forked and spreading over and over again.
Social media platforms are no longer limited to user interaction when it comes to satisfying their seemingly bottomless desire for personal information. Once you have your data, you are no longer restricted to misuse it for marketing purposes. Data is sold and purchased to data brokers, popularized via LLM output, and is placed in a much wider variety of applications. Often, it is against the wishes expressed without user informed consent.
Research like that brought Incogni's ranking of social media privacy is a way to get a roadmap for laying this rapidly evolving landscape and choosing routes that lead to a brighter future.
The complete analysis (including public datasets) can be found here.
[ad_2]
Source link
