Siri can call Google now. Where does confidential work go?

Apple just did something I did not expect from the privacy company. It let the new Siri call Google.

When you ask Siri a challenging question, it doesn’t always provide an answer directly on the phone. Instead, it can route requests to Google’s Gemini and to Apple’s Private Cloud Compute, which now operates partly in third-party data centers.

This design choice makes sense for a general assistant, as complex queries often require more computing power than a phone can provide. However, it raises a critical issue that many professionals may not have considered yet.

If Apple is sending its most difficult questions to a partner cloud, the assumption that “it runs on my device” is no longer a given. This shift means products now have to prove where the work actually runs, not just promise it.

For most users, this change may not be a concern. However, for certain professionals, it could be a dealbreaker. Consider a therapist taking session notes, a lawyer drafting privileged documents, an inventor discussing unfiled ideas, or a screenwriter safeguarding a script. For them, the default convenience could become a significant liability.

This year, the trend has shifted negatively, as the two leading AI assistants began training on consumer conversations by default in 2025. In 2024, regulatory bodies, including HHS and the American Bar Association, cautioned that sharing confidential work with third-party AI could be considered a disclosure.

This leads to an important question: Where does sensitive work that cannot leave the device actually go? Currently, it mostly ends up in the same clouds as everything else.

This gap is what I am focusing on building toward.