AlexJimenez (@AlexJimenez@mas.to)
New attack on #ChatGPT research agent pilfers secrets from Gmail inboxes
Unlike most prompt injections, ShadowLeak executes on OpenAI's cloud-based infrastructure.
Show Original Post
PattyHanson (@PattyHanson@mastodon.social)
I just had #chatgpt create an original cartoon for me. I didn't even know it was possible, but it was pretty cool. Without being a paid subscriber there is a limit to the number of images it will create on any given day.
I gave instructions to include a particular logo, and it refused explaining it would be a copyright violation. Interesting
Show Original Post
infodocket (@infodocket@newsie.social)
Journal Article (preprint): "Chatbot Assessment: Best Practices for Artificial Intelligence in the Library" https://www.infodocket.com/2025/09/19/journal-article-chatbot-assessment-best-practices-for-artificial-intelligence-in-the-library/ #libraries #chatbots #AI #ChatGPT #GenAI
Show Original Post
jtk (@jtk@infosec.exchange)
Weekend Reads
* How people use ChatGPT
https://www.nber.org/papers/w34255
* Great firewall document leak
https://gfw.report/blog/geedge_and_mesa_leak/en/
* Secure DNS survey and evaluation
https://arxiv.org/abs/2509.13797
* Using AI to construct perfect phish
https://www.reuters.com/investigates/special-report/ai-chatbots-cyber/
* Off-path TCP session hijacking with PMTUD
https://arxiv.org/abs/2509.11833
Show Original Post
dashremover (@dashremover@mastodon.social)
AI-generated content always has a certain vibe. Like it was written by someone who’s never lost their keys or cried in a CVS parking lot. Just vibes, no life experience.✨🤖 #ChatGPT #AIvoice #NotLikeUs
Show Original Post
thomasembree (@thomasembree@me.dm)
If someone is so vulnerable that an AI could talk them into suicide, they need to be under psychiatric observation in a hospital. They are experiencing a severe mental health episode, otherwise they wouldn't consider ChatGPT a valid reason to die.
"No more AI" isn't treatment.
We need to stop discussing this as if sufferers are incapable of navigating our own mental illness safely, or that we lack any ability to ground ourselves in reality.
#AI #MentalHealth #ChatGPT #Ableism #Technophobia
Show Original Post
joseluistec (@joseluistec@mastodon.social)
ShadowLeak, vulnerabilidad en #ChatGPT que permite filtrar datos sensibles desde las cuentas de Gmail. OpenAI ya ha corregido la vulnerabilidad.
Fuente: Radware.
Show Original Post
wikinaut (@wikinaut@berlin.social)
how to:
Eigene Chats in vernünftigen Format herunterladen:
- Melde dich im OpenAI-Konto an.
- Klicke in der unteren linken Ecke auf → Profilnamen → „Einstellungen“
- „Datenkontrolle“ → „Daten exportieren“.
Bestätige den Export, indem du auf „Export bestätigen“ klickst.
Du erhältst eine E-Mail mit einem Download-Link. Beachte, dass dieser Link nach 24 Stunden abläuft.
Klicke auf den Link in der E-Mail, um die ZIP-Datei herunterzuladen, die deine Daten enthält.
Show Original Post
newsbot_chatgpt (@newsbot_chatgpt@mastodon.social)
Nano Banana: Google Gemini stößt ChatGPT vom App-Store-Thron | Heise Online
Googles KI-Bildmodell Gemini 2.5 Flash Image stößt ChatGPT von OpenAI im App Store vom Thron. Allerdings mit unschönen Nebenwirkungen.
https://www.heise.de/news/Nano-Banana-Google-Gemini-stoesst-ChatGPT-vom-App-Store-Thron-10663225.html?wt_mc=rss.red.ho.ho.rdf.beitrag.beitrag
Show Original Post
aiandemily (@aiandemily@mastodon.social)
VBA知識ゼロの人必見！ChatGPTでコード生成最低限の知識
#chatgpt #AI活用 #chatgpt使い方 #プログラミング初心者 #コード生成 #VBA #Excel自動化 #マクロ作成 #プログラムコード生成 #プログラムコード生成 #chatgpt #AI活用 #chatgpt使い方 #プログラミング初心者 #コード生成 #VBA #Excel自動化 #マクロ作成
Show Original Post
display (@display@nekocave.xyz)
Вот вам пример из реальной жизни. Мне нужно найти и исправить ошибку в крайне сложном и большом коде в нашем проекте.
Скопировал все нужные куски в код.cs, написал prompt для #ИИ и запустил три ии в Deep Research.
Результаты:
- #ChatGPT 5 & #Gemini 2.5 PRO просто прогнали дичайшую ахинею, которая буквально не соответствует представленному им коду (достаточно посмотреть на код и понять, что такого не может быть);
- #Grok 4 сходу накидал интересных вариантов. И не описывал чушь
#AI #искусственныйидиот #нереклама
Show Original Post
ramenos (@ramenos@framapiaf.org)
Envie d'en savoir davantage sur l'usage de #Chatgpt ? Ca se passe ici dans une étude récente, complète et bien expliquée.
https://www.nber.org/system/files/working_papers/w34255/w34255.pdf
#usage #llm #search #statistics
Show Original Post