Fact Check: Meta claims new features are built in the most privacy-oriented way possible.

Fact Check: "Meta claims new features are built in the most privacy-oriented way possible."

What We Know

Meta recently announced new generative AI features, emphasizing that they were developed with a strong focus on user privacy. According to their official statement, they implemented privacy mitigations through a comprehensive review process and external engagements (Privacy Matters: Meta's Generative AI Features). Notably, Meta claims that their AI models, including Llama 2, were trained without using any user data from their platforms. The training data consisted primarily of publicly available information and licensed data, with specific filters to exclude sources that typically share personal information (Privacy Matters: Meta's Generative AI Features).

Furthermore, Meta has stated that they do not use the content of private messages for training their AI models. They also provide users with options to delete information shared during interactions with their AI features, thereby enhancing user control over their data (Privacy Matters: Meta's Generative AI Features).

Analysis

The claim that Meta is building new features in a privacy-oriented manner is supported by their detailed disclosures regarding data usage and privacy safeguards. The transparency about not using private user data for training models and the implementation of user controls, such as the ability to delete chat information, lend credibility to their assertion (Privacy Matters: Meta's Generative AI Features).

However, skepticism exists regarding Meta's commitment to privacy. Critics, including privacy experts, have expressed concerns about the company's track record with user data and privacy violations. For instance, there are ongoing discussions about Meta's introduction of ads in WhatsApp, which has raised alarms among privacy advocates who question whether such monetization efforts might compromise user privacy (WhatsApp ads are here – and privacy experts are worried).

While Meta's current claims are bolstered by their stated practices and transparency measures, the historical context of their privacy issues, including a record $5 billion fine in 2019 for privacy violations (Meta Faces Stiff Privacy Review as FTC Claims New Violations), suggests that their assurances should be viewed with caution. The reliability of their claims can be further scrutinized in light of past behaviors, which may lead to skepticism about their current privacy promises.

Conclusion

The claim that "Meta claims new features are built in the most privacy-oriented way possible" is True. Meta has outlined specific measures taken to ensure user privacy in the development of their new generative AI features, including not using private user data for training and providing users with control over their interactions. However, the historical context of privacy violations and ongoing scrutiny from experts necessitates a cautious approach to fully trusting these claims.

Fact Check: Meta claims new features are built in the most privacy-oriented way possible.