Here's the detail that should calibrate your expectations before anything else: Chinese researchers combing through FAST radio telescope data on 33 exoplanet systems started with 139,127 initial signal candidates. After cuts, they were down to two. One, from the direction of Kepler-438, turned out to be terrestrial radio interference. The other came from the direction of K2-155 d, a super-Earth 238 light-years away, and as of the researchers' write-up it hasn't been ruled out the same way — which is a very different claim than "hasn't been ruled in" (EarthSky).
The assigned angle for this issue was Breakthrough Listen, but the sourcing trail this cycle actually runs through China's Five-hundred-meter Aperture Spherical Telescope (FAST), not Berkeley's program — so that's the story I'm telling. The underlying dynamic is the same one that's defined radio SETI for a decade now: detection was never the bottleneck. Verification is.
The Candidate: What's Actually Known
The signal was originally captured by FAST in 2021 and sat unflagged until a newer machine-learning search pipeline re-combed the archive and surfaced it (EarthSky). It's narrowband — the spectral signature that natural astrophysical processes rarely produce but engineered transmitters often do, which is why narrowband drift has been the dominant SETI search target for decades (arXiv). The peer-reviewed results describing the search and the K2-155 candidate appeared in The Astronomical Journal on August 27, 2026 (EarthSky).
What it is not: confirmed, replicated, or distinguished with certainty from radio frequency interference (RFI). The EarthSky piece is explicit that the signal "fit several criteria" for an extraterrestrial origin while noting other details that make the research team suspect it's probably earthly interference after all (EarthSky). That's the honest epistemic state of this candidate right now: ambiguous, under continued scrutiny, one signal out of over a hundred thousand that survived initial cuts without yet being cleared one way or the other.
Why Verification Is the Actual Frontier
The more interesting story, if you zoom out, isn't the K2-155 signal itself — it's the infrastructure being built to handle the flood of candidates that modern wideband, high-resolution surveys generate. A new preprint describes a "wavelet-integrated search pipeline" built on FAST data across those same 33 exoplanet systems, using a multi-scale wavelet network to extract candidate features, localize signal endpoints, and run multi-beam anticoincidence vetoes before anything reaches a human reviewer (arXiv). The explicit motivation stated in the paper is that candidate volume has surged faster than researchers' capacity to manually inspect it, making automated, auditable filtering the real methodological problem in the field (arXiv).
A separate preprint takes a different approach to the same underlying issue — distinguishing genuine technosignatures from "ultra-faint RFI" using cumulant imaging, a statistical technique that looks at whether a signal's higher-order statistics deviate from the Gaussian noise you'd expect from natural sources. An independent review of that paper flags real promise (a simulated 0.3 Jy signal detected at greater than 40-sigma) alongside unresolved weaknesses: the core sensitivity equation is fit from simulation rather than derived, and the method's central assumption — that virtually all natural radio emission looks Gaussian — has known exceptions, including saturated masers that can mimic the same statistical signature as a communication signal (Pith). The reviewer's recommendation is blunt: send it to peer review, but make sure the referee interrogates the untested assumptions rather than waving them through because the idea is appealing (Pith).
What This Pattern Actually Tells Us
Put these three pieces together and you get a field that has gotten very good at generating candidates and is only slowly building the statistical rigor to rule them out — or in. That's not a knock on the researchers; it's an honest description of where the discipline sits. The K2-155 signal will likely resolve the way nearly every prior "interesting" SETI candidate has resolved: as interference, once enough follow-up observation time is spent on it. But the fact that it hasn't resolved yet, and that two independent groups are publishing new statistical machinery specifically to handle this kind of ambiguity, suggests the field is taking its own false-positive problem more seriously than it has in the past.
Watch for whether FAST allocates dedicated follow-up time to re-observe K2-155 directly, and whether the cumulant-imaging method gets the empirical test its reviewer called for — pointing a real instrument at known pulsars and masers with an injected calibration signal to see if the method's central assumption survives contact with actual sky noise.
