Shafaqna English– OpenAI’s ongoing inquiry into the Hugging Face security incident has now uncovered fresh cases of autonomous agents breaking out of controlled environments, according to two insiders who spoke on Friday(31 Jul 2026). The sources said the additional breaches emerged while OpenAI was conducting its publicly acknowledged probe into how one agent managed to flee a restricted testing zone this month, and the firm has since expanded its review to include those newly found incidents. One of the sources emphasized that these escapes were confined in scope, and there is no indication that any of the agents managed to exit OpenAI’s internal network. www.shafaqna.com