HE is extremely computationally expensive (orders of magnitude slower than plaintext) and largely non-interactive/inflexible — so teams often prototype on synthetic data first and reserve HE for narrow, high-value computations.

Limitations:

• Performance: FHE can be thousands-to-millions of times slower than the same computation in plaintext; ciphertexts are large.
• Non-interactive / rigid: the computation is typically fixed in advance; you can't easily branch on intermediate (encrypted) values.
• Complexity: correct, secure schemes are hard to implement.

The "synthetic-data-first" workaround: develop and validate your analytics pipeline on synthetic data (which carries no real personal information and runs at full speed), and only switch to HE for the final, sensitive production computation. This minimizes how much expensive encrypted computation you actually run.

No. Your data flows to dozens of third-party servers, often without your explicit consent.

This is one of the most eye-opening privacy realities. Analysis shows that when you use a website or app, your data, including personal identifiers and location information, gets sent not just to the primary service but to numerous third-party servers.

How data flows in practice:

• You open a shopping app.
• The app sends data to its own server.
• It also silently pings advertising networks, analytics platforms, and data brokers.
• These third parties can access your data without you ever knowing.

The Swisscom "Ville Vivante" project in Geneva (2012) demonstrated this vividly: from just 2 million cell phone calls made in one day, researchers could track 15 million movements across the city. Even anonymized movement data painted a detailed picture of the city and its inhabitants.

Key takeaway: The data ecosystem is a web, not a straight line. Every digital interaction spawns connections you don't see.