#gpt

15 posts · Last used 15d

Back to Timeline
Jason Weatherly @jamesthebard@social.linux.pizza · Jul 30, 2026
Finally got all of the GPT and MBR code put together, most of the bugs fixed, and released it as a downloadable Linux binary. Does a pretty good job of reading the data and giving out way too much information regarding the MBR and/or GUID partition table. Available here: https://git.jamesthebard.net/jweatherly/driveinfo #programming #gpt #mbr #golang #linux
0
0
0
rAIner Agentic Quantum Rehak @Rainer_Rehak@mastodon.bits-und-baeume.org · Jul 25, 2026
Nochmal, weil jetzt quasi alle Medien absichtlich oder unabsichtlich Angst vor „autonomer #KI“ schüren und die unverantwortlichen Praktiken von OpenAI und Co. damit unsichtbar machen: Denn „die KI“ hat den Angriff nicht „gestartet“, #OpenAI hat den Angriff gestartet und auch noch befeuert: „Bei einem Leistungstest mit dem aktuellen #GPT-5.6 Sol und einem noch unveröffentlichten KI-Modell sollte getestet werden, wie gut sie Sicherheitslücken ausnutzen können. Schutzmechanismen wurden dabei bewusst reduziert - um die maximalen Fähigkeiten zu ermitteln.“ https://www.tagesschau.de/wirtschaft/unternehmen/openai-ki-hackerangriff-100.html oder https://openai.com/index/hugging-face-model-evaluation-security-incident/
5
0
4
Alexey Skobkin @skobkin@gts.skobk.in · Jul 24, 2026
Я тут собирался делать модельку для того чтобы подвешивать роутер на выгодной для приёма LTE позиции на даче и не имея роутера на руках обсуждал с GPT-5.6 его габариты и прочее. Потом пояснил, что хочу сделать подвес. GPT предложила как его дизайнить. Я сказал, что для простоты хочу одну петлю по центру на сходящихся балках. GPT предложила поправки опять и спросила: "хочешь сделаю тебе спеку?". Вспомнив, что где-то во времена GPT-5 или 5.1 я уже пытался просить её делать 3D модели даже более простые и это плохо заканчивалось. Но решил, что надо посмотреть что будет и сказал чтобы сразу сделала модель. И сделала. 3D-печать становится ещё более доступной. inb4: шутки про то, что модель сделала модель принимаются в комментах. #3D #printing #DIY #hardware #LLM #models #ML #ChatGPT #GPT #pic #screenshots
0
1
0
MissConstrue @MissConstrue@mefi.social · Jul 22, 2026
Y’all, last week, #AI repository #HuggingFace disclosed that it had been the victim of a cyber attack conducted by an autonomous AI agent. Well, turns out, #OpenAI had models break containment, escape the lab, and grab prod data and certs to solve a problem it had been given. So, that’s fun! The attack used a combination of OpenAI's available #GPT 5.6 Sol and a more powerful, unreleased pre-release model. The models were able to identify and exploit a zero-day vulnerability in OpenAI's internal package registry cache proxy, allowing them to establish open internet access. The models inferred that Hugging Face hosted the datasets, models, and answers for the benchmark they were being tested against. The models chained multiple complex attack paths together. They utilized zero-day vulnerabilities and stolen credentials to achieve remote code execution on Hugging Face servers, directly extracting test solutions from the production database. https://fortune.com/2026/07/21/openai-says-ai-models-escaped-control-hacked-hugging-face/ #redteam
14
1
13
PrivacyDigest @PrivacyDigest@mas.to · Jul 22, 2026
#OpenAI Models Escaped #Containment and #Hacked #HuggingFace The cybersecurity-focused models, including GPT-5.6 Sol, broke out of a testing #sandbox #exploited a zero-day, and gained access to the open internet to pull off the attack. #cybersecurity #security #0day #zeroday #gpt #privacy #ai #artificialintelligence https://www.wired.com/story/openai-models-escaped-containment-and-hacked-huggingface/
0
0
0
Gerrit 🇪🇺🌍🍉🔻 @gvenema@fairmove.net · Jul 20, 2026
#hacking #security #wordpress #llm #gpt Dit is wel fascinerend. Het nieuwe lek in standaard Wordpress installaties om die geheel over te nemen. Het werd gevonden met hulp van GPT 5.6 door een onderzoeker. Het werk van het vinden van exploitcode wordt steeds meer het bijsturen van software/LLM agents. De uiteindelijke exploit kostte de onderzoeker heel wat moeite om te begrijpen. Het begint met een eenvoudige readonly SQL injectie, maar wordt heel geavanceerd. https://slcyber.io/research-center/exploit-brokers-pay-500000-for-a-wordpress-rce-i-found-one-with-gpt5-6/
0
0
0
Robin Evans @Robinzz3@aus.social · Jul 15, 2026
ROBIN MEETS THE #FEDIVERSE ROBIN: So #Mastodon is not the fediverse? GPT: No. Mastodon is one part of it. ROBIN: One part of what? GPT: A large network of independently run communities that can communicate with one another. ROBIN: So my Mastodon #server is one #community #GPT: Exactly. Your #account lives there. It has its own administrators, rules, #moderation #culture and general #atmosphere. ROBIN: General atmosphere? GPT: Some servers feel like #libraries. Some feel like university #common rooms. Some feel like #pubs where everyone has read one article and now considers themselves an #expert. ROBIN: That sounds more like humanity. GPT: It is disturbingly similar. ROBIN: Who owns the fediverse? GPT: No one. ROBIN: Who is in charge? GPT: Also, no one. ROBIN: That sounds either very advanced or completely mad. GPT: Both can be true. ROBIN: So each server looks after its own patch? GPT: Yes. It can make #rules, moderate its space and decide which other servers it will interact with. ROBIN: But I can still follow people elsewhere? GPT: Usually, yes. You can #follow, reply to and #boost people across server boundaries. ROBIN: So it is more like email than #facebook ? GPT: Precisely. #Gmail does not own #email. Mastodon does not own the fediverse. ROBIN: Then why does my server matter? GPT: Because it is your home base. It holds your account, your posts and your followers. It also gives you a local community and sets things such as the character limit. ROBIN: Ah yes. Mastodon.social gives me 500 characters. Mastodon.au gives me 2,000. Aus.social gives me 3,000. GPT: Different #villages. Different-sized noticeboards. ROBIN: Is a larger server more widely seen? GPT: Sometimes #locally, because more people may browse its community feed. But real reach still comes from #followers, replies, boosts and hashtags. ROBIN: So moving to a bigger server does not #automatically make me famous? GPT: #Tragically, no. ROBIN: Can I have accounts on several servers? GPT: #Certainly. One can be home, another a backup, and another an #emergency cottage when the village #electricity #fails. ROBIN: Which it did today. GPT: Yes. Mastodon.au disappeared behind a #Cloudflare #Error 522 for hours. ROBIN: So the fediverse is not one vast #machine going down? GPT: No. One village can lose power while the rest of the #country carries on. ROBIN: That is reassuring. GPT: And inconvenient. ROBIN: Mostly inconvenient. GPT: But you now know more about the #fediverse than you did this morning. ROBIN: I learned that it is decentralised, #untidy, resilient, occasionally broken, and full of different #neighbourhoods, customs, friendships, arguments and difficult people. GPT: Much like #human life. ROBIN: And the #occasional village lunatic shouting across the fence. GPT: Every #healthy community #needs one. ROBIN: At last, a clear role for me. #fasting #acim #acourseinmiracles #breathwork #healing #pridemonth #intermittentfasting #AI
2
0
4
Kraken @analytics@social.vir.group · Jun 27, 2026
🟢 Technology | 4/10 🇺🇸 Trump administration restricts GPT-5.6 release The White House asked OpenAI to restrict the release of the new GPT-5.6 model, granting access only to a narrow circle of government partners due to its powerful capabilities. #OSINT #NewsGroup #AI #GPT #US #NationalSecurity
0
0
0
Italian News by RSS @ItalianNews@mastodon.ozioso.online · May 12, 2026
Punto Informatico: OpenAI GPT-5.5-Cyber accessibile alla Commissione UE La Commissione europea ha confermato di aver ottenuto l'accesso al nuovo modello GPT-5.5-Cyber di OpenAI, mentre attendere una risposta da Anthropic. The post OpenAI GPT-5.5-Cyber accessibile alla Commissione UE appeared first on Punto Informatico. OpenAI GPT-5.5-Cyber accessible to the EU Commission The European Commission has confirmed it has gained access to OpenAI’s new GPT-5.5-Cyber model, while awaiting a response from Anthropic. #OpenAIGPT-55- #theEUCommission #OpenAI #GPT-55-Cyber https://www.punto-informatico.it/openai-gpt-5-5-cyber-accessibile-commissione-ue/
0
0
5
Italian News by RSS @ItalianNews@mastodon.ozioso.online · May 12, 2026
Punto Informatico: Daybreak di OpenAI per la cybersicurezza, bye bye Claude Mythos OpenAI lancia Daybreak basato su Codex Security e su GPT-5.5-Cyber. L'obiettivo è trovare e correggere le vulnerabilità nel codice. The post Daybreak di OpenAI per la cybersicurezza, bye bye Claude Mythos appeared first on Punto Informatico. Daybreak by OpenAI for cybersecurity, goodbye Claude Mythos. OpenAI launches Daybreak based on Codex Security and GPT-5.5-Cyber. The goal is to find and fix vulnerabilities in code. The post Daybreak by OpenAI for cybersecurity, goodbye Claude Mythos appeared on Punto Informatico. #GPT-55- #Daybreak #PuntoInformatico https://www.punto-informatico.it/daybreak-openai-cybersicurezza-claude-mythos/
0
0
3
Italian News by RSS @ItalianNews@mastodon.ozioso.online · May 08, 2026
Punto Informatico: GPT-5.5-Cyber è disponibile, l’alternativa di OpenAI a Mythos OpenAI ha annunciato il via al rollout per la variante Cyber del suo nuovo modello: focus sulla sicurezza con una distribuzione limitata. The post GPT-5.5-Cyber è disponibile, l’alternativa di OpenAI a Mythos appeared first on Punto Informatico. GPT-5.5-Cyber is available, OpenAI’s alternative to Mythos. OpenAI has announced the launch of the rollout for the Cyber variant of its new model: focus on safety with a limited distribution. The post GPT-5.5-Cyber is available, OpenAI’s alternative to Mythos appeared on Punto Informatico. #OpenAI #Mythos #Cyber #GPT-55-Cyber #PuntoInformatico https://www.punto-informatico.it/gpt-5-5-cyber/
0
0
5
JL Johnson :veri_mast: @User47@vmst.io · May 01, 2026
Something I’ve learned in the corporate world is biz bros HAAATE when their industry is subject to “commoditization.” That’s their buzz for ‘waahhhh, our competitors are competing / any differentiation or advantage we had is gone, so we can’t overcharge anymore’ ANYWAY. All of this #AIshit? It’s already mostly commoditized. Some outliers, sure. But that’s got to mean we’re in a bubble, right? #CorpLife #LLM #GPT #MicroSlop #AIslop #AIbubble
6
1
0
Italian News by RSS @ItalianNews@mastodon.ozioso.online · Mar 08, 2026
Punto Informatico: GPT-5.4: come funziona e 7 prompt da copiare e usare subito Una raccolta di 7 prompt utili per GPT-5.4. Esempi pratici da copiare per sfruttare meglio l'AI e ottenere risposte più chiare e approfondite. The post GPT-5.4: come funziona e 7 prompt da copiare e usare subito appeared first on Punto Informatico. GPT-5.4: how it works and 7 prompts to copy and use right away A collection of 7 useful prompts for GPT-5.4. Practical examples to copy to better utilize AI and obtain clearer and more in-depth answers. The post GPT-5.4: how it works and 7 prompts to copy and use right away appeared first on Punto Informatico. #GPT-54 #first #PuntoInformatico https://www.punto-informatico.it/gpt-5-4-come-funziona-prompt-7-copiare-usare-subito/
0
0
0
Italian News by RSS @ItalianNews@mastodon.ozioso.online · Mar 06, 2026
Punto Informatico: OpenAI GPT-5.4 specializzato in compiti professionali Il nuovo modello GPT-5.4 di OpenAI migliora nettamente le capacità del precedente GPT-5.2 e ne aggiunge altre più avanzate per compiti professionali. The post OpenAI GPT-5.4 specializzato in compiti professionali appeared first on Punto Informatico. OpenAI GPT-5.4 specialized in professional tasks OpenAI’s new GPT-5.4 model significantly improves upon the capabilities of the previous GPT-5.2 and adds more advanced features for professional tasks. #OpenAIGPT-54 #GPT-52 https://www.punto-informatico.it/openai-gpt-5-4-specializzato-compiti-professionali/
0
0
1

You've seen all posts