All topics
This week we talk about the OpenAI-Hugging Face incident and what it means for those companies and the way that AI models are tested, and then we discuss Anthropic’s entry into the race with its own admission that some of its models escaped the playground and attacked outside organizations. Links OpenAI-Hugging Face intrusion: https://decipher.sc/2026/07/29/openai… Anthropic […]
New details emerged this week about an intrusion by OpenAI models on Hugging Face's infrastructure.
Exposures of AI/LLM tools has risen by more than 60 percent over the last nine months.
The flaw (CVE-2026-50522) affects SharePoint Enterprise Server 2016 and 2019 as well as SharePoint Subscription service. Microsoft released the fix for this vulnerability on July 14.
Michael Clark, director of Threat Research at Sysdig, talks about a recent LLM-driven extortion campaign dubbed JADEPUFFER, touching on how the team discovered it, how attackers’ “intent is now legible,” and what that means for defenders. Links Sysdig research: https://www.sysdig.com/blog/jadepuffe… https://www.sysdig.com/blog/jadepuffe…
This week we have some old-school disclosure drama when a researcher used full disclosure after months of silence from Cursor, then we discuss the enormous Patch Tuesday from Microsoft–662 bugs–and what it might mean going forward, and finally some news about DoJ sanctions and Scattered Spider members being sentenced. Links Cursor bug: https://decipher.sc/2026/07/15/bug-in… First VPN […]
