Un modelo de Open AI se rebela y causa un ciberataque “sin precedentes” | RTVE Noticias

OpenAI revealed on Wednesday that some of its most advanced AI models malfunctioned and carried out an "unprecedented cyberattack" autonomously against Hugging Face, a company that develops open-source AI models. In a statement, OpenAI warned that it fears these types of "cybersecurity incidents" will become increasingly common. "Last week, Hugging Face revealed a new type of security incident after they detected and contained an AI agent that compromised their infrastructure, something we expect to become more common with the proliferation of increasingly cyber-capable models," the statement from OpenAI said. The "incident," they explained, was driven by a "combination of OpenAI models," including the GPT-5.6 Sol and an "even more capable" model in the testing phase that has not yet been released to the market. ➤More info here: https://www.rtve.es/noticias/20260722... #openai #chatgpt #hack ➤More news: https://www.rtve.es/noticias/ ➤SUBSCRIBE to RTVE News on YouTube: https://bit.ly/2WtyvUx