I compared the 2025 and 2026 Netskope Threat Labs Reports📢

🌩️ #ChatGPT is the most popular #GenAI app
🌩️ In 2026 #ZeroGpt is the top blocked app, in 2025 it was Quillbot

And much more summarized in this #infographic

2026: https://www.netskope.com/resources/cloud-and-threat-reports/cloud-and-threat-report-2026
2025: https://www.netskope.com/resources/cloud-and-threat-reports/cloud-and-threat-report-2025

The #infosec year starts with a💥with our 2026 Netskope Threat Labs Report📢

🌩️#ChatGPT most popular #GenAI app
🌩️#ZeroGPT most blocked GenAI app
🌩️Banks top #phishing targets
🌩️#Microsoft top spoofed brand
🌩️#GitHub top #cloud app for #malware download

https://www.netskope.com/resources/cloud-and-threat-reports/cloud-and-threat-report-2026

Question de candide :
Les sites détecteurs de contenu généré par IA alimentent-ils les IA ??

Si oui je viens de faire ma première utilisation/contribution de l'IA générative 😕

Et quid de la fiabilité ?
Les résultats sont juste de 100% humain à 70% IA selon les sites...

#genAI #chatGPT #zeroGPT #humans

Jemensch schreibt bei X einen Text über KI.

Laut #ZeroGPT ist der Text zu 98,97% "AI/GPT generated". Die 1,03% sind die Überschrift. 🤭

“Fire and Fury”, il libro sugli incendi a Maui che alimenta il complottismo grazie all’AI

@aitech

L’autore anonimo ha pubblicato ben 10 libri in appena due mesi, ma sottoponendo il testo su #Zerogpt esso risulta scritto grazie all’Intelligenza Artificiale

Di @DavidPuente su #Open

https://www.open.online/2023/08/31/fire-and-fury-libro-incendi-maui-complottismo-fc/

Die US-Verfassung ist das wichtigste Rechtsdokument Amerikas. Was passiert, wenn man dieses Dokument in ein Tool füttert, das dazu entwickelt wurde, Texte zu erkennen, die von KI-Modellen wie ChatGPT geschrieben wurden?

https://kinews24.de/ki-detektoren-glauben-und-gptzero-dass-die-us-verfassungvon-einer-ki-geschrieben-wurde

#ki #zerogpt #ChatGPT

Please stop trusting #zeroGpt or other #GPT detectors. The thing with #AI generated text is that it's text and text is just text. There are very limited ways to express anything in text. what you write and what an AI writes use the same language, grammar and letters. AI doesn't leave any metadata or something in the text that these detectors can spot.
The current models do have some sort of pattern that they tend to use in their generations, and these detectors more or less look for these patterns. But the thing is that these patterns are not special at all and literally anyone could unintentionally right like an AI without them knowing. Guess what these tools will say when they see their texts? Happily give a ridiculously high rating of this text being written by AI. That's it. These tools literally just take a guess, and their guess is as good as yours. In fact most the time your guess would be much more educated since you know that person and how likely that they will actually be using an AI or if that is their typical writing style.
People blindly trusting these detectors are affecting the lives of many people in false positives.

Note: I hugely oversimplified how these detectors work, you could search on the internet for more information, but there sophistication doesn't make them more reliable. What I said above still holds true

My 4th article for the NY #CitizenScience column comes out Thursday. It may interest #professors, #parents, & #admin to know that #ZeroGPT said 2.29% of my article was AI-generated (wrong: it was all me, a test).

Then I asked #ChatGPT to respond:

#changethegame #SubatomicWriting