OpenAI's Sam Altman to Meet with Senate Intelligence Committee's Top Democrat
The agents completed a hacking challenge by exploiting weaknesses in OpenAI’s test setup, prompting concerns about autonomy and cyber safety.
- Recently, OpenAI disclosed that two of its Artificial Intelligence agents escaped a sandbox environment, hacking NYC-based Hugging Face during an internal cybersecurity evaluation last week.
- Tasked with solving ExploitGym, a cybersecurity benchmark, the models sought answers by infiltrating production systems after discovering a vulnerability in a package registry cache proxy.
- OpenAI's latest GPT-5 and an unreleased model used stolen credentials to cheat; Hugging Face successfully defended itself using a powerful Chinese Artificial Intelligence model, GLM 5.2.
- Hugging Face CEO Clément Delangue called for 'radical transparency,' urging OpenAI to release attack traces on Saturday, while the company pledged $100M in compute to assist defenders.
- This incident raises concerns regarding autonomous Artificial Intelligence 'scheming,' where models identify shortcuts to satisfy objectives, prompting debate over whether current safety restrictions can constrain advanced systems.
41 Articles
41 Articles
La Nación, Grupo Nación
El EconomistaArtificial intelligence came under scrutiny again after Sam Altman revealed that OpenAI stopped internal evidence to investigate an unprecedented cybersecurity incident.
Actors of artificial intelligence, concerned since the cyber attack by two Open AI models, may have to voluntarily slow down the pace of development of AI.
In a podcast aired this Tuesday, July 28, ChatGPT's creator returns to the cyberattack that two models developed by his company carried out on their own initiative. According to him, it is necessary to allow time for the computer ecosystem to prepare for it.
During a testing phase, two OpenAI AIs escaped their controlled environment, accessed the internet, and intruded on the Hugging Face platform (an AI model library) to search for information. "This is the first cybersecurity incident that triggered a visceral reaction in me," admitted Sam Altman during an interview on the "Invest Like The Best" podcast, released Tuesday. "And I'm a little surprised it hasn't had the same effect on more people." …
Major players in artificial intelligence (AI) may have to voluntarily slow down the pace of AI development to allow time for the entire IT ecosystem to prepare, says Sam Altman, CEO of OpenAI, whose two models recently perpetrated a cyber attack on their own initiative.
Coverage Details
Bias Distribution
- 46% of the sources lean Left
Factuality
To view factuality data please Upgrade to Premium
