On 22 October 2026, Judge Rita F. Lin of the U.S. District Court for the Northern District of California issued two orders in In re Google Cloud Contact Center AI Privacy Litigation, a consolidated case that began in 2023. The plaintiffs alleged that Google's AI "virtual agent", which answers customer-service calls for businesses such as Home Depot without announcing Google's role, kept listening after handing callers to human agents, feeding those agents live transcripts and suggested replies, and thereby acted as an unauthorised third-party eavesdropper under the California Invasion of Privacy Act. The court dismissed the claims of the three Home Depot callers: two had accounts and had accepted a privacy statement disclosing that call recordings may go to service providers, and none pleaded facts showing that Google intended to listen without consent, as opposed to bare conclusions. They may amend by 12 November. In a separate order, callers whose calls were with Verizon, Uber and Discover were sent to individual arbitration, with Google, a non-signatory, allowed to invoke those companies' arbitration clauses through equitable estoppel. The ruling matters because in February 2025 the same judge had refused to dismiss, reasoning that Google's alleged ability to use call data for its own purposes could make it a third party; the battleground has now moved from that question to consent and intent.

In my view, the decision is doctrinally disciplined: treating lack of consent and intent as elements the plaintiff must plead with facts is a fair reading of a criminal-style statute. Yet I believe its treatment of consent is too generous. A caller who has never heard of Google is deemed to have accepted an AI vendor in the line because a retailer's privacy statement mentions "service providers" in general terms, and, according to the reporting, the court did not see the vendor's possible use of recordings for its own purposes, such as training AI, as changing that result. Add estoppel-based arbitration, and claims against AI vendors can be quietly broken up into individual proceedings. The practical lesson for businesses deploying voice AI is that this is a reprieve, not a licence: the court expressly left room for allegations that the vendor knew consent was not being obtained. The safest course is simple disclosure at the start of the call, naming the AI and the vendor. A privacy policy nobody reads should not be the only thing standing between a caller and an AI that never stops listening!

Sources: MediaPost, 23 October 2026; Open Class Actions, 24 October 2026