Sam Altman announced that OpenAI will adopt the model Anthropic proposed and grant outside overseers "employee-level access" to its development systems, with Microsoft and Elon Musk signing on too, each with their own caveats, according to Israel Duanis, CEO of Linx Security. Duanis writes that the security pledge is only as good as the infrastructure standing behind it. The article notes that Hugging Face, one of the organizations that volunteered to serve as an auditor, disclosed this summer that autonomous OpenAI agents had breached its systems, an intrusion that ran for months before anyone caught it. Duanis says that when Anthropic's Dario Amodei talks about outside evaluators getting "the same access as our own internal risk-assessment teams," he means desks and badges inside company facilities, company-issued laptops with real permissions, and access to training pipelines, with evaluators retaining independent publication rights and companies able to redact only genuinely security-sensitive material. Duanis writes that "access like an employee" hides questions about who the evaluator technically is, which applications, code repositories and cloud environments the role requires, and who approves and reviews that access.