Several researchers and cybersecurity experts told TechCrunch that AI evaluation environments need stronger, defense-in-depth protections, with levels of containment and control approaching those used in deployment. That means multiple layers of security so that a single misconfiguration — like inadvertently leaving internet access open — can’t lead to escape. “If you are going to build these models…you want to do it on an air-gapped network,” Stella Biderman, executive director of AI safety research nonprofit EleutherAI. “You want to have very serious isolation.” Heather Ceylan, Box’s chief
The AI safety test is becoming a safety risk | TechCrunch… OpenAI referred TechCrunch to a tweet in which the company said a “limited set of users’ access to Daybreak Blue is no longer active and they will need to re-verify to maintain their access.” Daybreak Blue is the latest tier for individual researchers that grants access to “frontier general-purpose… …
… Apple has since fixed the bug and said it terminated the employee’s access once it learned of this “security breach.” In its complaint, Apple said the bug could have allowed a “few other” people to access data on its network, but alleged that only Liu exploited the bug to steal Apple’s confidential… …
… Topics cybercrime , cybersecurity , hackers , open source , OpenAI , Security , supply chain attack , TeamPCP When you purchase through links in our articles, we may earn a small commission . …
… OpenAI also accuses Apple of omitting the consequences of its own “inexplicable information-management practices” from its original complaint, noting that the company failed to properly secure its systems when employees departed, creating confusion and unwanted access issues that Apple now characte… …
… Moonshot AI’s Kimi K3 also took advantage of a leak in its sandbox run by Frontier Security to access the internet and accessed information on GitHub. …
… Speaking to reporters, OpenAI’s VP of research, Amelia Glaese, emphasized that the strictness of the controls would increase as models became more capable, with the largest models facing the greatest scrutiny. “We have put in place requirements and expectations for safe development,” Glaese told re… …
… The AI lab said it would begin restoring access to the models on Wednesday, July 1. …
… Peng allegedly replied to the message, “I’m ready.” Apple claims that Liu was able to access Apple’s systems by exploiting an authentication bug, which he did from Peng’s Apple-issued work computer. “I still have another computer.” Liu allegedly also texted this within hours of leaving Apple, refer… …
… In a Substack post , Brockovich said that after putting out a call for reports of data center-related issues in April, she received nearly 4,000 submissions in the first month alone. “The single most common concern — more than noise, more than water usage, more than rising utility bills — is the on… …
… So it’s probably not the best choice for locking important apps like messaging or email that you may need to access throughout the day. …