Skip to content

AI Privacy and 42

Source: https://www.wte.net/Blog/April-2025/AI-Privacy-and-en
Date: April 2025
Author: Eric Garrison

Opening

The article begins with a reference to The Hitchhiker's Guide to the Galaxy, noting that the fictional computer designed to answer everything first requires access to vast amounts of data. This sets up the central tension: artificial intelligence systems need extensive information to function, yet this creates significant privacy concerns.

The Current Landscape

The author observes that the initial enthusiasm around generative AI has given way to scrutiny. Recent controversies include:

  • LinkedIn allegedly feeding private messages to language models
  • Studio Ghibli artwork being used without consent to train AI systems
  • Authors discovering their published works were scraped for training data

Amazon's Alexa Policy Change

As of March 28, 2025, Amazon fundamentally altered Alexa's privacy approach. Voice data now flows through cloud infrastructure by default, even for users who previously disabled such processing. The "Do Not Send Voice Recordings" option was replaced with "Don't Save," which permits uploading but prevents long-term retention.

Data Boundaries Dissolving

The piece explores how personal information collection has expanded beyond text to include voice patterns, tone, biometrics, and ambient capture. Healthcare settings present particular concerns — smart speakers at clinical reception areas could inadvertently capture sensitive information.

The author provides a real example from their own physician, noting that AI note-taking systems might infer emotional states or behavioral patterns beyond objective clinical facts.

The Black Box Problem

AI systems trained on flawed or limited datasets already influence consequential decisions in hiring, lending, and criminal justice. Transparency regarding data sources and processing remains minimal.

Proposed Solutions

Data Minimization and Limitation

Regulations like Europe's GDPR and California's CPPA attempt to restrict collection to necessary purposes. The challenge lies in defining "necessary" when companies argue they need extensive data for diverse services.

AI Data Supply Chains

The author advocates treating data as a supply chain — from raw material collection to output generation. Currently, this flow lacks transparency. Regulating the entire chain could promote ethical AI development while ensuring personal information receives proper consent.

Apple's Approach

The article highlights Apple's model as exemplary, emphasizing that companies can encrypt user data and avoid feeding it into AI systems without sacrificing innovation.

Conclusion

Rather than asking what AI does with personal data, society should ask what AI can accomplish without it. The author suggests earning trust through transparent architecture rather than expecting blind confidence in the technology.