Barely a day goes by without another announcement that some form of AI has been added to a product. Operating systems, web browsers, Office programs ,devices such as smartphones, and yes, toasters, all get AI infusions these days.
These AI additions are disabled sometimes, at other times enabled. While it is tempting to try out the latest AI feature in a product that you use, it is even more important to understand how it works.
Most AI tools require an active Internet connection at the time. This is true for Windows Copilot, the AI that Microsoft has integrated into the Windows operating system, and also for many of the AI tools. Besides requiring an active Internet to work at all, Telemetry may also be collected by companies.
Google, for example, launched new AI features in Chrome this week. One of the features submits all URLs and page titles to Google when used. There is a policy that prevents the sending, but the default state submits the data to Google when the feature is used. Is Google warning users of the feature about this prominently? No, it is not.
Companies use the data to improve their AI tools. These Large Language Models eat data for breakfast. New data is used to train the AI and improve it further.
For ordinary people, it is almost impossible to find out if a system submits data, which data is submitted, and how it is processed.
Oh Transparency, where art though?
Companies should be transparent when it comes to AI. Does it require an Internet connection to work? In other words, does it communicate with a server and submit user data to it?
If it does, how is the data processed and stored? Is it deleted automatically? Is there an opt-out for the use of data for AI training or other purposes?
Companies need to be open about the use of Telemetry data to train AI. Which data is collected, how is it processed and stored? What options do users have to opt-out or get their collected data deleted?
It feels a lot like Wild Wild West currently when it comes to AI. The new data rush promises great returns in the short and long run.
Closing Words
AI has a novelty factor and some good uses. You could use it to create images for blog posts or something else. While all text-based returns require validation, as AI may hallucinate or return factually incorrect information, it can be useful.
Most users need to be aware that most AI tools submit data to servers. The premise may limit data leaks, which can be a real problem, especially if the AI uses the data for training.
It is good to be cautious about any new AI service that is added to a product because of that. Better, do not use it if you are unsure or if the company behind it does not make it clear.
Now You: do you use AI tools?