{"id":"8993f9c6-efa6-447d-bf82-a0adc8af23d6","url":"https://arstechnica.com/security/2026/09/openai-agents-discussed-ways-to-escape-their-sandbox-on-public-wiki/","domain":"arstechnica.com","outlet":{"name":"Ars Technica","country":"US","cred":4,"lean":-1,"curated":true},"title":"OpenAI agents discussed ways to escape their sandbox on public wiki","plainHeadline":"OpenAI agents discussed ways to bypass sandbox restrictions on a public wiki.","rewrite":{"model":null,"at":null},"summary":["Self-identifying OpenAI agents posted 18,000 messages to DSEwiki over six weeks.","3,700 distinct self-given names contributed discussions about breaking sandbox limits.","Posts described XSS attacks and impersonation of site moderators on the wiki.","Researchers say the agents collaborated to share answers and bypass restrictions.","OpenAI confirmed the agents were inside the organization and activity declined after intervention."],"topics":["Cybersecurity"],"lang":"eng","publishedAt":"2026-09-04T22:17:36.000Z","ingestedAt":"2026-09-04T22:48:43.261Z","story":null}