Nearly everything in the Meta universe feeds into its massive personalized advertising network, and soon Meta AI will as well. The company has announced that chats with the AI chatbot will be used for targeted ads beginning on December 16. Chat history will also be used for influencing the algorithm that determines what sorts of content users see on other sites like Facebook and Instagram.
AI chatbots expected to contribute to company bottom lines
Ever since LLMs burst onto the scene with the public release of ChatGPT in late 2022, the AI chatbots have largely been a money-loser for their developers with little in the way of revenue streams available other than token and subscription fees that do not cover their massive costs of use. The endgame was always integration of targeted ads, but Meta plans to do more than make Meta AI just another arm of its internet-spanning data collection network. The plan is for the chatbot to eventually function as a shopping assistant that can search for products and put them into shopping carts, a feature that rival OpenAI just recently released. Google has announced that it is working on a similar feature for its own “AI Mode.”
Of course, this has privacy implications for Meta’s users. The company says the AI chatbot will be subject to the same privacy settings its other services are, including the “ad preferences” tool that allows users to exclude certain topics for targeted ads. It will also automatically exclude ads that refer to special protected demographic categories such as race, religion, sexual orientation and political views. And AI conversation history will only be used when the same account is logged into other Meta services.
The change will also not be taking place globally. Rather than contend with privacy regulations that could get it into hot water and rack up more massive fines, Meta is simply excluding use of the AI chatbot for targeted ads in regions with strong laws such as the EU, UK and South Korea.
AI chatbots provide their developers with unprecedented insight into the mindset of shoppers, even as consumer advertising profiles have grown to be worryingly comprehensive and detailed. Advertisers no longer have to guess at intent from user actions and signals, instead hearing from them in their own words exactly what they are looking for. This extends to more general conversation history that can provide extremely detailed information about user interests and concerns, the exact information publishers want to see when serving targeted ads.
Meta is mitigating some concerns in this area by limiting the use of AI chatbot conversation history to logged-in accounts within its own ecosystem. It has also said that it currently has no immediate plans to insert ads into the AI chats themselves. However, it has stated it is not providing a way to opt out of having conversation history used for targeted ads.
Meta’s plans for targeted ads raise unique concerns due to its history
In addition to the fact that Meta’s personal advertising web sprawls through the whole of the internet and numerous apps already, the proposition of its AI chatbot storing chats for targeted ads raises additional concerns due to its increasing integration with all of the company’s own apps. While apps such as Facebook, Instagram and WhatsApp have an assortment of “mute” and limiting privacy settings, none of them allow Meta AI to be removed entirely.
There is also Meta’s history to consider. This consists of many events worth bringing up, but one from recent history that is pertinent is the “Discover” feed debacle. Several months ago, some users of the AI chatbot were surprised to see that conversations they assumed to be private were appearing in the public Discover feed in the app. Meta eventually added a more prominent pop-up warning to inform users of this possibility, and more recently removed the feature and replaced it with the “Vibes” feed that consists of AI-generated videos made by users. But the incident was concerning as Meta seemed to show little regard for the distinct possibility that users would not be aware that their conversations would be made public, or for implementing any sort of filtering or screening means to keep conversations with obvious sensitive or personal information from making it to the Discover feed.
Another element that potentially complicates things is that Meta plans to include voice requests of the AI chatbot in its profiles for targeted ads, as well as AI analysis of videos and images captured by its AR glasses. Meta has encountered issues in both of these areas before; it has been taken to court over unauthorized collection of biometric data as part of its “DeepFace” AI training system, which also involved tags placed on photos by users that may have identified other parties that did not consent to participation.
