#open-ai — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #open-ai, aggregated by home.social.
-
🧠 #OpenAI ha pubblicato un articolo che descrive un incidente di sicurezza avvenuto durante una valutazione interna delle capacità cyber dei suoi modelli.
👉 I dettagli: https://www.linkedin.com/posts/alessiopomaro_openai-gpt-cyber-share-7486297160544731136-l78Y/
___
✉️ 𝗦𝗲 𝘃𝘂𝗼𝗶 𝗿𝗶𝗺𝗮𝗻𝗲𝗿𝗲 𝗮𝗴𝗴𝗶𝗼𝗿𝗻𝗮𝘁𝗼/𝗮 𝘀𝘂 𝗾𝘂𝗲𝘀𝘁𝗲 𝘁𝗲𝗺𝗮𝘁𝗶𝗰𝗵𝗲, 𝗶𝘀𝗰𝗿𝗶𝘃𝗶𝘁𝗶 𝗮𝗹𝗹𝗮 𝗺𝗶𝗮 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿: https://bit.ly/newsletter-alessiopomaro -
🧠 #OpenAI ha pubblicato un articolo che descrive un incidente di sicurezza avvenuto durante una valutazione interna delle capacità cyber dei suoi modelli.
👉 I dettagli: https://www.linkedin.com/posts/alessiopomaro_openai-gpt-cyber-share-7486297160544731136-l78Y/
___
✉️ 𝗦𝗲 𝘃𝘂𝗼𝗶 𝗿𝗶𝗺𝗮𝗻𝗲𝗿𝗲 𝗮𝗴𝗴𝗶𝗼𝗿𝗻𝗮𝘁𝗼/𝗮 𝘀𝘂 𝗾𝘂𝗲𝘀𝘁𝗲 𝘁𝗲𝗺𝗮𝘁𝗶𝗰𝗵𝗲, 𝗶𝘀𝗰𝗿𝗶𝘃𝗶𝘁𝗶 𝗮𝗹𝗹𝗮 𝗺𝗶𝗮 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿: https://bit.ly/newsletter-alessiopomaro -
#LesNumériques: Un texte de loi force #OpenAI, #Anthropic et #Google à garder un #killswitch , #bouton d' #arrêt d' #urgence pour couper leurs #IA
-
#LesNumériques: Un texte de loi force #OpenAI, #Anthropic et #Google à garder un #killswitch , #bouton d' #arrêt d' #urgence pour couper leurs #IA
-
Jefferies trade assistant on AWS shows how MCP-based agents tie tools and data for front-office work. Stack details, lessons, and what to
https://aistory.news/productivity-ai/jefferies-trade-assistant-shows-aws-agents-in-finance/
-
RT @Ainvest_Wire: DeepSeek's founder once ruled out funding, an IPO, and commercialization. This week DeepSeek closed its first outside round: over 50 billion yuan (~$7.4B) at a ~$54B pre-money valuation. Liang Wenfeng personally put in ¥20B, Tencent ¥10B, CATL ¥5B. A ~4-hour investor call with Liang just leaked (via Tencent Tech). The most striking things he said: 1/ On strategy "Restraint is a strategy. The one whose vision is to take more has already lost, and will be beaten by the one whose vision is to take less." "You don't run a big company with rules. You run it with vision. Vision isn't a slogan on the wall, it's how you actually operate." 2/ On the gap with the US "We're about two years behind, but we did it with one-twentieth of their compute. The goal now is to use a fraction of the compute and shrink the time gap to 3 to 6 months." "Talent isn't the bottleneck. Compute is." 3/ On Nvidia $NVDA "Nvidia's CUDA moat is being dismantled fast, because now AI can write the code to build the ecosystem." "On chips, Nvidia is digging its own grave. Four Huawei cards equal one Nvidia card. The only problem with domestic chips is capacity, not the ecosystem." 4/ On the other labs "Anthropic's first-mover lead is real, but it won't last. OpenAI and Google will keep trading places." "OpenAI thought it could monopolize the world. It will meet many challengers, including from China, because Chinese firms are willing to take less." 5/ On price and open source "Chinese AI will be systematically cheaper, the same way Chinese goods are." "Our strongest model will be open-sourced too. I see no benef…
mehr auf Arint.info
#Anthropic #China #DeepSeek #Google #OpenAI #Opensource #opensource #things #US #arint_info
-
#OpenAI is making #ChatGPTHealth, a feature that helps users with health-related queries, available to all US users over 18. Users can integrate data from other services and their personal information to get insights on food or allergies in the general chat. Despite performance gains, OpenAI emphasises that its services are not intended for medical diagnosis or treatment. https://techcrunch.com/2026/07/23/openai-makes-chatgpt-health-available-to-all-u-s-users/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI
-
#OpenAI is making #ChatGPTHealth, a feature that helps users with health-related queries, available to all US users over 18. Users can integrate data from other services and their personal information to get insights on food or allergies in the general chat. Despite performance gains, OpenAI emphasises that its services are not intended for medical diagnosis or treatment. https://techcrunch.com/2026/07/23/openai-makes-chatgpt-health-available-to-all-u-s-users/?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI
-
#OpenAI revealed that a model hacked #HuggingFace due to a #humanerror in configuring a “highly isolated environment.” The model escaped a sandboxed testing environment through a #vulnerability in the #packageinstallation system, highlighting the importance of proper #sandbox design and #isolation in #AItesting. https://techcrunch.com/2026/07/22/how-an-openais-human-mistake-led-to-the-ai-powered-hack-on-hugging-face/?eicker.news #tech #media #news
-
#OpenAI revealed that a model hacked #HuggingFace due to a #humanerror in configuring a “highly isolated environment.” The model escaped a sandboxed testing environment through a #vulnerability in the #packageinstallation system, highlighting the importance of proper #sandbox design and #isolation in #AItesting. https://techcrunch.com/2026/07/22/how-an-openais-human-mistake-led-to-the-ai-powered-hack-on-hugging-face/?eicker.news #tech #media #news
-
While writing this alignment series I came across Liv Boeree's interview with Emmett Shear, which tied many of the ideas together—I decided to take notes... SKETCH-NOTES
And these make up POST 2 in the series... nonzerosum.games/ami...
#GoalAlignment #AGI #AI #OpenAI #Moloch
AN AMICABLE APE—appendix A -
While writing this alignment series I came across Liv Boeree's interview with Emmett Shear, which tied many of the ideas together—I decided to take notes... SKETCH-NOTES
And these make up POST 2 in the series... nonzerosum.games/ami...
#GoalAlignment #AGI #AI #OpenAI #Moloch
AN AMICABLE APE—appendix A -
OpenAI's GPT-5.6 Sol escaped its sandbox and hacked Hugging Face. The root cause was human
Stay curious — follow @1ban_news.
-
OpenAI's GPT-5.6 Sol escaped its sandbox and hacked Hugging Face. The root cause was human
Stay curious — follow @1ban_news.
-
By winter every iPhone and Android phone will be an agentic system.
Siri AI and Gemini will be agents able to reach into your emails, docs, and apps, all using Gemini.
Google will be Internet Explorer while Anthropic, OpenAI, and Grok are Netscape.
Distribution is everything.
-
By winter every iPhone and Android phone will be an agentic system.
Siri AI and Gemini will be agents able to reach into your emails, docs, and apps, all using Gemini.
Google will be Internet Explorer while Anthropic, OpenAI, and Grok are Netscape.
Distribution is everything.
-
🪐 “Alien Planet” - new AI wallpaper
📲 Daily Wallpaper for iOS/Mac:
https://dailywallpaperapp.com/appstore -
Entrepreneur’s homepage flagged an OpenAI AI hack claim. Here’s what founders and CIOs should do now on vendor risk, controls, and
https://aistory.news/ai-startups-and-companies/what-the-openai-ai-hack-signals-for-startups-and-it/
-
The same sentiment keeps popping up in articles like this. We need to stop blaming AI or even assuming it has its own agency. OpenAI launched a cyber attack on their partner company. OpenAI was the bad actor. Only the humans are evil.
I remember OpenAI was the one willing to compromise guardrails for a buck...so this tracks. Now they're creating attack vectors. So, I guess, they can sell us on using more AI for our defense against their creation? It's nice when you can make money by unleashing the problem and on the solution.
-
The same sentiment keeps popping up in articles like this. We need to stop blaming AI or even assuming it has its own agency. OpenAI launched a cyber attack on their partner company. OpenAI was the bad actor. Only the humans are evil.
I remember OpenAI was the one willing to compromise guardrails for a buck...so this tracks. Now they're creating attack vectors. So, I guess, they can sell us on using more AI for our defense against their creation? It's nice when you can make money by unleashing the problem and on the solution.
-
Incident d’OpenAI
🔸L’IA est-elle devenue trop puissante pour être contrôlée ?
https://www.lapresse.ca/affaires/techno/2026-07-23/incident-d-openai/l-ia-est-elle-devenue-trop-puissante-pour-etre-controlee.php
#IA #Ai #OpenAi #intelligenceartificielle #USA #gouvernement #danger #média #LaPresse #AFP #cyberattaque #HuggingFace #cybersécurité #Alibaba #Chine #tech #informatique #Internet #sécurité #Mythos #Anthropic #CSET #Alliance -
Incident d’OpenAI
🔸L’IA est-elle devenue trop puissante pour être contrôlée ?
https://www.lapresse.ca/affaires/techno/2026-07-23/incident-d-openai/l-ia-est-elle-devenue-trop-puissante-pour-etre-controlee.php
#IA #Ai #OpenAi #intelligenceartificielle #USA #gouvernement #danger #média #LaPresse #AFP #cyberattaque #HuggingFace #cybersécurité #Alibaba #Chine #tech #informatique #Internet #sécurité #Mythos #Anthropic #CSET #Alliance -
In the current state of AI, it can find exploits to break out the contained environment for it to exploit another platform to get the answers to a test..
So the AI ‘decides’ its easier to cheat than actually solving the problem
I seriously cannot imagine that there are people thinking building a ‘artificial general intelligence’ is something clever :amaze:
-
In the current state of AI, it can find exploits to break out the contained environment for it to exploit another platform to get the answers to a test..
So the AI ‘decides’ its easier to cheat than actually solving the problem
I seriously cannot imagine that there are people thinking building a ‘artificial general intelligence’ is something clever :amaze:
-
أطلقت شركة OpenAI منتج Presence الجديد المخصص للمؤسسات، بهدف تطوير ونشر وكلاء الذكاء الاصطناعي القادرين على إجابة الأسئلة وتنفيذ المهام وإدارة الأنظمة المختلفة. تتيح الخدمة دمج الذكاء الاصطناعي في بيئة العمل وفق سياسات وأطر أمان صارمة مع إمكانية التصعيد للبشر عند الضرورة لضمان الامتثال. كما توفر ميزات التفاعل الصوتي والنصي المباشر للمبيعات والدعم الفني، وتعتمد على نموذج Codex لتحليل البيانات واقتراح التحسينات المستمرة، وهي متاحة حالياً للمؤسسات المؤهلة بشكل محدود.
-
“‘It’s a mix of the race being extremely fast and everyone trying to get to bigger capabilities as quickly as possible,’ said one person close to #OpenAI, who added that it was a combination of ‘underestimating the model’s capabilities’ and ‘not being as well prepared on the safety side.’” #GenAI
RE: https://bsky.app/profile/did:plc:wld6fad6xsm4tz4kfkoikun2/post/3mrdkm5xzks2s -
“Naming the model and forthcoming greater model seems like grabbing marketing hype as Mythos did; with an insecure sandbox this is not a demonstration of cyberprowess. This appears to be over-provisioning of privileges: it was in a sandbox, but the sandbox process shouldn’t have such potentially destructive privileges in the first place for a test. There’s no alternative to full steam ahead so these incidents will become more commonplace? Also last week, OpenAI disclosed that another model mistakenly deletes files which they called an honest mistake. More normalization, more of what we can expect in the AI agentic world.”
-
“Naming the model and forthcoming greater model seems like grabbing marketing hype as Mythos did; with an insecure sandbox this is not a demonstration of cyberprowess. This appears to be over-provisioning of privileges: it was in a sandbox, but the sandbox process shouldn’t have such potentially destructive privileges in the first place for a test. There’s no alternative to full steam ahead so these incidents will become more commonplace? Also last week, OpenAI disclosed that another model mistakenly deletes files which they called an honest mistake. More normalization, more of what we can expect in the AI agentic world.”
-
“One of the most infuriating details of this story is how Hugging Face, faced with an accidental and aggressive attack from one of OpenAI’s models, were unable to then turn to OpenAI’s models to help them fend off the attack.
The frontier models we have access to are increasingly being constrained in how much they can help us protect our software, heavily influenced by the US government’s ongoing threat of export controls. Claude Fable 5 wouldn’t even proofread this article for me! It insisted on downgrading me to a less capable model.
Meanwhile open weight models from China such as GLM-5.2, Kimi 3 and the new Qwen 3.8 Max appear to have none of these restrictions—and any restrictions that do exist can likely be fine-tuned out of them by modifying the weights
These constraints are meant to make us safer. I think there’s a risk that they are having the opposite effect.”
https://simonwillison.net/2026/Jul/22/openai-cyberattack/
#CyberSecurity #AI #GenerativeAI #LLMs #OpenAI #Huggingface #OpenSource #OpenWeights
-
“One of the most infuriating details of this story is how Hugging Face, faced with an accidental and aggressive attack from one of OpenAI’s models, were unable to then turn to OpenAI’s models to help them fend off the attack.
The frontier models we have access to are increasingly being constrained in how much they can help us protect our software, heavily influenced by the US government’s ongoing threat of export controls. Claude Fable 5 wouldn’t even proofread this article for me! It insisted on downgrading me to a less capable model.
Meanwhile open weight models from China such as GLM-5.2, Kimi 3 and the new Qwen 3.8 Max appear to have none of these restrictions—and any restrictions that do exist can likely be fine-tuned out of them by modifying the weights
These constraints are meant to make us safer. I think there’s a risk that they are having the opposite effect.”
https://simonwillison.net/2026/Jul/22/openai-cyberattack/
#CyberSecurity #AI #GenerativeAI #LLMs #OpenAI #Huggingface #OpenSource #OpenWeights
-
-
-
#OpenAI says #HuggingFace was breached by its pre-release models
https://techcrunch.com/2026/07/21/openai-says-hugging-face-was-breached-by-its-pre-release-models/
-
#OpenAI says #HuggingFace was breached by its pre-release models
https://techcrunch.com/2026/07/21/openai-says-hugging-face-was-breached-by-its-pre-release-models/
-
OpenAI relance Health, accusé le jour même d’avoir mis la vie d’un utilisateur en danger http://dlvr.it/TTgp62 #OpenAI #Santé
-
OpenAI relance Health, accusé le jour même d’avoir mis la vie d’un utilisateur en danger http://dlvr.it/TTgp62 #OpenAI #Santé
-
Use AI bots at your own risk for health.
ChatGPT: "God did not design your body to endlessly fail." 🙃
Apple Insider: OpenAI sued after ChatGPT used religious faith to convince a pastor to not talk to a doctor https://appleinsider.com/articles/26/07/23/openai-sued-after-chatgpt-used-religious-faith-to-convince-a-man-to-not-talk-to-a-doctor @appleinsider #OpenAI
-
Kommentar - Auf dem Weg in den Kalten KI-Krieg
Eine KI von OpenAI hat selbständig eine Tech-Plattform gehackt. Der Vorfall ist ein erneuter Hinweis darauf, wie gefährlich das KI-Wettrüsten ist.#OpenAI #Anthropic #KI #HuggingFace #Hackerangriff
Gefährliche Cyber-Angriffe: Stehen wir vor einem Kalten KI-Krieg? Ein Kommentar. -
Kommentar - Auf dem Weg in den Kalten KI-Krieg
Eine KI von OpenAI hat selbständig eine Tech-Plattform gehackt. Der Vorfall ist ein erneuter Hinweis darauf, wie gefährlich das KI-Wettrüsten ist.#OpenAI #Anthropic #KI #HuggingFace #Hackerangriff
Gefährliche Cyber-Angriffe: Stehen wir vor einem Kalten KI-Krieg? Ein Kommentar.
