How to use WhatsApp AI safely and responsibly

The meta, which owns Facebook, Instagram and WhatsApp, implemented its artificial intelligence (AI) in the instant messenger, expanding the functionality of the application. The technology, which had been available in other countries since April 2024, took a little longer to arrive in Brazil due to impositions from the National Data Protection Authority (ANPD).

WhatsApp's AI is based on advanced language models, such as LLAMA (acronym for Large Language Model Meta AI), trained with large amounts of textual data to understand and generate natural language effectively. “ The meta's AI is able to answer questions, give recommendations, research news about subjects of our interest on the web without leaving the app and generating images and small GIFs for sharing”, explains Pierre dos Santos, AI analyst at Leste Telecom.

“However, the tool is still in beta phase, so it contains many errors in its use. This will be improved over time, and AI may even have its revised mode of use, as it has many opportunities for adding new services, including those related to accessibility”, he adds.

Girl or villain?Depends on use

With so much debate about the use of artificial intelligence, which has already been proven to be behind practices such as fake news and deepfake, many people are afraid to have the AI of the goal available on WhatsApp, without even the possibility of deactivating the feature. “The meta informed that the content of the conversations with the AI could be used to train the artificial intelligence algorithms, but does not associate this content with the personal information of the users”, reassures Pierre.

Although it has not publicly disclosed how the data will be used for ad targeting, in addition to the focus on AI training, the continuous use of the tool may have an impact on receiving advertisements and long-term advertisements. Data collection, which is a common practice in the technology market, can be used for personalization of ads, audience segmentation and behavior forecasting, for example.

“ However, my expectation is that the goal will prioritize the user's privacy and consent, using AI in an ethical and transparent way to benefit both users and advertisers in line with our legislation”, says the analyst.

Even if AI does not have access to private conversations on WhatsApp and user data remains protected by messenger encryption, according to AI documentation, messages shared with the tool can be used to provide relevant answers to you or to improve this technology. “Sofore, do not send messages containing information you do not want to share with AI. At least, we can delete messages sent to AI by typing /reset-all-ais into the” conversation, warns the analyst.

Use sparingly

Pierre also says that AI is a powerful tool that can be useful in many contexts. However, it is essential to use responsibly and care, always thinking about the security and privacy of personal data. For this, he shares some basic but valuable tips:

• Use AI as an auxiliary tool, not as a substitute for critical thinking;
• Use AI for tasks you consider safe and without risk to your privacy, avoiding sharing personal or sensitive information with AI in conversation;
• Avoid using AI to make important decisions;
• Research only on topics of general interest, avoiding sensitive or controversial subjects.

“It is true that every day it is increasingly difficult to identify whether content was generated by AI, but there are some signs that can help you distrust: unknown or doubtful source; content too good to be true; lack of information about authorship; artificial language; generic content and without originality; and lack of emotion and subjectivity,” concludes the expert.