The TypeSafe "leak of all customer noul outputs" was a joke: the gist lists 0.00 to 1.00
publicado (OK do Fabricio 25/10 BRT).
On Friday 23 October 2026 at 22:49 BRT (01:49 UTC on the 24th), TypeSafe CEO Diogo Almeida posted: “we became aware of a massive unauthorized leak of ALL customer noul outputs”. Then, “in effort to maximize transparency”, he linked the data.
The linked gist is a single file, outputs.json, with 101 numbers: 0.00, 0.01, … 1.00. A noul is Jev’s yes/no question type, which returns one probability between 0 and 1. The “leak” is the full list of values a noul can take, rounded to two decimals, so it contains no customer data. Almeida’s replies in the thread (“sorry we’re calibrated”, “can’t even sue us”) are part of the joke. The post passed 5,800 likes.
Some coverage took it at face value. Startup Fortune ran “TypeSafe AI admits a leak exposed every customer’s Jev model output” and speculated about an IDOR bug; Jev Daily and several X accounts also described it as a real incident. We found no security notice from TypeSafe and no customer report backed by anything other than the 101-number gist.
For people using Jev, nothing changes: outputs are still probabilities over options you define. More on the model: Jev.