What happened to Loccus.ai

This page used to be a head-to-head comparison. It is not one any more, because there is nothing on the other side of it. What follows is a disposition: what Loccus was, what happened to it, where the technology went, and what an acquisition changes for the person who was about to buy.

What Loccus.ai was

Loccus.ai was an audio-first specialist in synthetic speech detection — one of the few companies that decided voice deserved its own product rather than a modality slot inside a broader deepfake platform. That is the same bet we have made, which is part of why this page existed at all.

The claim that travelled furthest, still live in acquisition-era material when we checked on 31 August 2026, is that Loccus “can identify AI voices with just 1.5 seconds of audio”. Short-clip performance is the hard and useful problem here: real evidence arrives as a nine-second voicemail, not a minute of studio speech, and most detectors degrade badly as the clip shortens. Putting a number on the minimum was a specific, checkable claim rather than a vague one.

Whether it held up we cannot say. We never ran a clip past Loccus, no independent evaluation covering both of us exists, and the product is no longer available to test.

What happened

Hiya announced the acquisition of Loccus.ai on 25 July 2024. The release names Alex Algard, Hiya’s CEO and Founder, and Manel Terraza, Loccus.ai’s CEO and Founder. The technology was rebranded as Hiya AI Voice Detection.

That is a sensible destination for a voice-detection specialist. Hiya operates where suspicious calls actually arrive — consumer apps, carrier deployments, a developer API — and detection is worth more in the path of a live call than behind a signup form. Distribution was the thing Loccus did not have and Hiya did.

Where the technology went

Into Hiya, and that is the comparison worth reading if you came here looking for a product to buy. Hiya publishes a per-dataset accuracy breakout across named public datasets, publishes hard operating limits, and — unusually for this market — does not sell voice cloning alongside detection. Those claims are dated and sourced on that page rather than here.

The developer documentation moved with it: docs.loccus.ai lands on developer.hiya.com, which is where the current reference lives.

What an acquisition means for a buyer

This is worth thinking about beyond one company, because detection is a fragmented field of specialists and the specialists keep getting bought.

When a product is acquired, three things stop being the thing you evaluated. The roadmap becomes someone else’s priority list, and a standalone audio detector inside a larger company now competes internally for engineering time. The API becomes someone else’s interface — different base URL, different auth, different rate limits, and a migration you did not schedule. The pricing becomes someone else’s commercial model, set against a different set of customers than the one you were in.

None of that is a criticism of the acquirer; it is a description of what changes hands. The buyer’s mistake is assuming the product continues unchanged because the technology did.

There is a real upside, usually left out by people making this point. A specialist inside a larger company gets distribution, a compliance apparatus it could not have afforded alone, and a longer runway than most detection startups ever raise. For someone who needs the technology to still exist in three years, being acquired is frequently the better outcome. What it is not is a neutral event you can ignore in an evaluation.

So, for any detector: ask who owns it, what the current model version is called, and what happens to your contract and your integration if the company changes hands. Those questions cost nothing.

Why this page still exists

Because people still search the name, and an answer is more useful than a redirect. Someone working from a vendor list written before July 2024, or auditing an integration built against docs.loccus.ai, deserves a page that says plainly what happened.

What this page deliberately does not do is compare Truthring against Loccus.ai. There is no live product on the other side and no honest version of that table. What we publish and what we do not is on the method, accuracy and limitations pages — and Truthring is pre-launch, with no measured figures yet.

Questions