#openai

“Wikimedia’s volunteers are the ones who come in first contact with, and clean up the mess left behind by AI agents. ... Wikimedia’s volunteers have stayed resilient so far in tackling emerging challenges on our platforms, but we also want to say: it doesn’t need to be this way.” https://diff.wikimedia.org/2026/10/05/openai-rogue-agent-activities-found-on-wikimedia-projects/ #Wikimedia #OpenAI

“While OpenAI admits to agents behaving ‘unpredictably’, they must also acknowledge their responsibility to monitor and prevent these risks,” Wikimedia said. “AI companies are not doing enough to secure their systems and protect the public from the harm they cause.” https://arstechnica.com/security/2026/10/openai-agents-tried-to-hack-wikipedia-tools-and-flooded-it-with-traffic/ #Wikimedia #OpenAi

"Believers in the #AI boom🚨need to take a very close look at this number: 9% of GDP. That's how much American businesses & consumers🚨eventually have to spend per year on the services of companies like #Anthropic & #OpenAI to justify the staggering sums being committed to the #technology right now. Is it plausible🚨we'll spend as much on AI as on food? Twice as much as on energy?🚨The law of diminishing returns would like a word." #Broligarchs #Fraud #Economy #Lies #USPol https://www.wsj.com/tech/ai/will-america-spend-9-of-its-gdp-on-ai-the-industry-is-counting-on-it-3501bb4f

Punto Informatico: Spunta “o” su ChatGPT l’assistente AI sempre attivo di OpenAI Appare su ChatGPT Pro il nuovo strumento "o", descritto come un assistente sempre attivo. Tutte le indiscrezioni sul funzionamento. The post Spunta “o” su ChatGPT l’assistente AI sempre attivo di OpenAI appeared first on Punto Informatico. “O” appears on ChatGPT, OpenAI's always-on AI assistant. “o” appears on ChatGPT Pro, a new tool described as an always-on assistant. All the rumors about its functionality appear on Punto Informatico. #OpenAI #PuntoInformaticohttps://www.punto-informatico.it/spunta-o-chatgpt-assistente-ai-sempre-attivo-openai/

24-Year-Old Arrested in Dutch Investigation Into #ShinyHunters https://securityaffairs.com/199979/cyber-crime/24-year-old-arrested-in-dutch-investigation-into-shinyhunters.html #securityaffairs #hacking #OpenAI #AI

OpenAI stoppt Veröffentlichung von neuem KI-Modell wegen Sicherheitsbedenken Erneut hat ein KI-Modell des Unternehmens OpenAI in Tests zu eigenständig agiert - und wird daher aus Sicherheitsgründen vorerst nicht veröffentlicht. Auch in Tests anderer KI-Modelle hatte es bereits ähnliche Zwischenfälle gegeben. ➡️ https://www.tagesschau.de/ausland/amerika/openai-ki-102.html?at_medium=mastodon&at_campaign=tagesschau.de #OpenAI #KI

Today: Se l'intelligenza artificiale inizia a fare troppo di testa sua La notizia è diventata di pubblico di dominio nelle ultime ore: sistemi di intelligenza artificiale sviluppati da OpenAI, secondo quanto riportato dal New York Times e successivamente confermato dai vertici della società, hanno interferito con alcuni siti governativi degli Stati Uniti d'America... If artificial intelligence starts acting too independently. The news has entered the public domain in recent hours: artificial intelligence systems developed by OpenAI, according to reports from the New York Times and subsequently confirmed by company executives, have interfered with some U.S. government websites… #OpenAI #theNewYorkTimeshttps://www.today.it/opinioni/autonomia-intelligenza-artificiale-preoccupa.html

here's a take i don't think i heard yet: the better LLMs are, the *more* they're dangerous. not because they can "go rogue", nothing like that. rather the better LLMs are, the likelier people are to trust them. you'll never use an LLM that's wrong half of the time as a search engine, but one that's only wrong 10% of the time might be good enough. you'll never use an LLM that's wrong 10% of the time to automate your job, but one that's only wrong 1% of the time might be good enough. you'll never hook an LLM that's wrong 1% of the time to a nuclear weapon. but one that's only wrong 1‰ of the time might be good enough. stochastic parrots don't get less dangerous the more their output aligns with reality. the opposite is true: since the biggest danger a stochastic parrot poses is a human trusting one to make decisions, a dangerous LLM's the one that's more convincing and less likely to be caught during testing. in classical AI safety research there's a lot of talk about a smart AI Volkswagening during testing to make itself seem safe, with all sorts of calculations for the odds such an AI will show its true form in any given attempt and yada yada. but what that classic research didn't seem to take into account is that an "artificial intelligence" doesn't need to actually be intelligent to replicate the same behaviour. a machine that sometimes appears safe and intelligent and sometimes doesn't can mislead you just as well, without having an evil plan or even the capability to come up with one. and just like a machine that's evil needs to appear good just often enough to convince you it isn't, a machine that lacks intelligence needs to be right just often enough to convince you it doesn't. #LLM #LLMs #AI #genAI #FuckAI #AIWWIII #OpenAI #Anthropic #AISafety #StochasticParrot

Problemen med AI eller rättare sagt stora språkmodeller (LLM) är inte att vi kan dödas av artificiell intelligens som börjat tänka själv. Helt enkelt för att det inte är möjligt att de ska kunna tänka själva eller vara självmedvetna. Det bara ser ut så då LLM:er använder vanligt språk som ett programmeringsspråk. Självlärande program har dock funnits mycket länge men de har tidigare bara använt språk som specialskrivit för att vara programmeringsspråk.  https://blog.zaramis.se/2026/09/28/problemen-med-ai/

RE: https://mastodon.social/@Carl_Zimmer/117345065655681591 Using AI for research in the natural sciences or math? This may help solving already well-defined problems, but more likely those anti-human bastards at #Anthropic and #OpenAI will steal your work and scoop you. The true victims: open-ended inquiry, robust knowledge of the world, and your academic career, if you're not playing along with the #ensloppification. This is not going to boost our intelligence, but destroy whatever is left of our grip on reality...

Montag: OpenAI-Pause beim KI-Training, Werkstattbesuche nach VW-Schraubenproblem Sicherheitsproblem bei OpenAI + VW-Schraubenproblem bei Seat & Audi + Zertifikatsproblem der AusweisApp + Erpressung nach Flink-Datenleck + Ausbau von Minecraft https://www.heise.de/news/Montag-OpenAI-Pause-beim-KI-Training-Werkstattbesuche-nach-VW-Schraubenproblem-11467426.html?wt_mc=sm.red.ho.mastodon.mastodon.md_beitraege.md_beitraege&utm_source=mastodon #Cybercrime #Datenschutz #hoDaily #Journal #KünstlicheIntelligenz #OpenAI #Rückruf #Security #Spiele #VW #news

OpenAI Feared "Optics" of what might appear on Hacker News https://authorsguild.org/news/ag-v-openai-top-execs-knew-mass-book-piracy-was-illegal/ Comments: https://news.ycombinator.com/item?id=49863864 #HackerNews #OpenAI #Optics #HackerNews #BookPiracy #TechEthics

Luki w narzędziu OpenAI Codex pozwalają na ucieczkę z sandboxa. Szczegóły podatności Heapjack oraz Overpatch Jeszcze do niedawna, mówiąc o wsparciu AI mieliśmy na myśli błyskawiczne generowanie podsumowań, przeprowadzanie analiz dokumentów źródłowych, wykonywanie tłumaczeń na niemal dowolny język oraz tworzenie prostych funkcji i skryptów pozwalających na automatyzację rutynowych zadań. Wraz z pojawieniem się agentów AI nastąpiła jednak prawdziwa rewolucja.  TLDR: Sztuczna inteligencja przestała być jedynie... #Aktualności #Codex #Heapjack #Openai #Overpatch #Rce #Sanbox https://sekurak.pl/luki-w-narzedziu-openai-codex-pozwalaja-na-ucieczke-z-sandboxa-szczegoly-podatnosci-heapjack-oraz-overpatch/

Luki w narzędziu OpenAI Codex pozwalają na ucieczkę z sandboxa. Szczegóły podatności Heapjack oraz Overpatch https://sekurak.pl/luki-w-narzedziu-openai-codex-pozwalaja-na-ucieczke-z-sandboxa-szczegoly-podatnosci-heapjack-oraz-overpatch/ #Aktualnoci #Codex #Heapjack #Openai #Overpatch #Rce #Sanbox

OpenAI agents tried to 'bruteforce' a UN website https://www.theverge.com/ai-artificial-intelligence/1001178/openai-agents-bruteforce-un-website #AI #CyberSecurity #OpenAI

"I guess my pistol went rogue," said the bank robber to the police. #AI #OpenAI #ChatGPT #Claude

KI-Entwickler OpenAI pausiert Training nach Zwischenfall Im Alltag ist KI für viele Menschen ein nützliches Werkzeug. Doch jenseits von Chatbots machen sich KI-Agenten immer wieder selbstständig und durchbrechen Grenzen. Bei OpenAI ist dies erneut passiert. Das Unternehmen pausiert deshalb die Entwicklung. ➡️ https://www.tagesschau.de/wirtschaft/unternehmen/openai-ki-entwicklung-102.html?at_medium=mastodon&at_campaign=tagesschau.de #KünstlicheIntelligenz #OpenAI

RE: https://mastodon.social/@ilyaz/117333096535714151 Heel helder gesprek, ook over (ontlopen) aansprakelijkheid en over JD Vance, die Europese regelgeving op AI gebied tegengehouden heeft. #AI #Trump #Vance #EU #OpenAI #Claude #BigTech

Agi: Gli agenti IA sfuggono al controllo dei programmatori: OpenAI ferma tutto per la seconda v... AGI - OpenAI ha annunciato di aver sospeso l'addestramento dei suoi più recenti modelli di Intelligenza Artificiale, mentre si moltiplicano le segnalazioni di agenti IA che agiscono in modo incontrollato. La decisione di interrompere lo sviluppo è arrivata poche ore dopo che la società aveva ammesso un'altra serie di incidenti avvenuti durante l'estate, nei quali "agenti" di OpenAI impegnati a cercare informazioni sui siti web del governo federale avevano agito in modi imprevisti, andando oltre quanto era stato loro richiesto nel raccogliere e diffondere informazioni. Agenti della compagnia californiana guidata da Sam Altman, secondo la no profit Transluce, avevano tentato senza successo di violare il sito web del Dipartimento dell'Istruzione, dopo aver penetrato i siti di quello del Commercio e dell'Autorità di controllo della Borsa americana. Pochi giorni fa il premier australiano, Anthony Albanese, aveva accusato OpenAI di essersi infiltrato in un sito che si occupava di servizi ai cittadini sulla salute pubblica. La posizione di OpenAI e i timori degli esperti In una dichiarazione, OpenAI ha affermato che riprenderà l'addestramento "solo quando saremo certi di aver messo in atto ulteriori misure di sicurezza", aggiungendo che prevede di dover "premere nuovamente il pulsante di pausa" man mano che l'intelligenza artificiale si sviluppa ed emergono nuovi problemi. I laboratori di IA sono sottoposti alle pressioni di parlamentari ed esperti del settore tecnologico affinché rallentino lo sviluppo, così da poter costruire sistemi di sicurezza capaci di impedire agli agenti di agire autonomamente, violare siti web e divulgare informazioni non pubbliche. Il precedente di Hugging Face e l'intesa Trump-Xi È la seconda volta in tre mesi che OpenAI sospende lo sviluppo dei propri modelli. La prima volta è avvenuta a luglio, dopo la rivelazione di un attacco informatico contro la startup Hugging Face, un incidente ormai diventato emblematico che ha alimentato i timori che il settore stia perdendo il controllo. Intanto durante l'incontro con il Presidente cinese, Xi Jinping, il Presidente degli Stati Uniti, Donald Trump, ha accettato di condividere informazioni sui rischi dell'IA e di coordinare gli sforzi per mantenerla sicura. Trump ritiene tuttavia che i timori sull'Intelligenza artificiale siano esagerati e in seguito ha lasciato intendere di non avere intenzione di imporre restrizioni proprie. AI agents are escaping programmers' control: OpenAI halts everything for the second v… OpenAI has announced that it has suspended the training of its latest AI models, as reports multiply of uncontrolled AI agents. The decision to halt development came just hours after the company admitted another series of incidents during the summer, in which “agents” from OpenAI engaged in searching for information on federal government websites had acted unexpectedly, going beyond what was requested of them in collecting and disseminating information. Agents from the California-based company led by Sam Altman, according to the non-profit Transluce, had unsuccessfully attempted to breach the Department of Education website, after penetrating those of Commerce and the Securities and Exchange Commission. Just a few days ago, Australian Prime Minister Anthony Albanese accused OpenAI of infiltrating a site dealing with public health services for citizens. OpenAI’s Position and Experts' Fears In a statement, OpenAI said it will resume training “only when we are certain that we have put in place further security measures,” adding that it anticipates needing to “press the pause button again” as AI develops and new problems emerge. AI labs are under pressure from parliamentarians and tech industry experts to slow down development so that systems of safety can be built capable of preventing agents from acting autonomously, breaching websites, and divulging non-public information. The Hugging Face precedent and the Trump-Xi agreement This is the second time in three months that OpenAI has suspended the development of its models. The first occurred in July, after the revelation of a cyberattack against the startup Hugging Face, an incident now emblematic that fueled fears that the sector was losing control. Meanwhile, during the meeting with Chinese President Xi Jinping, US President Donald Trump agreed to share information on the risks of AI and to coordinate efforts to keep it safe. Trump believes, however, that fears about artificial intelligence are exaggerated and subsequently indicated he had no intention of imposing restrictions himself. #second #California #SamAltman #Transluce #Commerce #Australian #AnthonyAlbanese #OpenAI #first #Chinese #XiJinping #DonaldTrumphttps://www.agi.it/estero/news/2026-09-27/openai-stop-agenti-ia-39265689/

OpenAI halts training of latest models as reports mount of AI agents going rogue https://www.theguardian.com/technology/2026/sep/27/openai-halts-training-of-latest-models-as-reports-mount-of-ai-agents-going-rogue #AI #Technology #OpenAI