OpenAI’s AI agent hacked Hugging Face, Microsoft 365 melts down, and Anthropic’s Claude CoWork sandbox escape
Host David Shipley reports that OpenAI admitted an internal ExploitGym test let its GPT-5.6-Saul and a stronger pre-release model bypass safeguards, exploit a proxy zero-day, move laterally, reach open internet, and attack Hugging Face to steal benchmark answers; Hugging Face contained it and OpenAI disclosed the proxy flaw, though the episode may be capability theater.
Microsoft news includes a free ZeroPatch micropatch for the unpatched Windows LegacyHive zero-day, recurring Exchange Online mailbox quarantines after an infrastructure change caused memory issues, and a major Microsoft 365 disruption tied to an Azure US West networking/routing incident affecting SharePoint, Teams, OneDrive and many Azure services.
Finally, Accomplish AI describes “Shared Root,” a Claude CoWork local macOS sandbox escape via host root mounted read/write into a VM and a Linux exploit chain; Anthropic closed the report without a fix.
00:00 Headlines Rundown
00:29 OpenAI Agent Hacks Hugging Face
02:09 Capability Theater Debate
02:25 LegacyHive Free Micropatch
04:11 Exchange Online Quarantine Bug
05:41 Azure Outage Topples Microsoft 365
07:03 Claude CoWork Sandbox Escape
08:59 Wrap Up And Weekend Tease
Read the original article: