A Basic LiteLLM Login Could Open the Control Room
An October 13 walkthrough shows how limited access to an AI gateway could reach its most sensitive controls. The flaws are patched—but the trust boundary deserves a closer look.
Tag Cybersecurity
The NSA Says “Never Trust” Is Coming for the Machines That Run Water and Power
Critical infrastructure depends on machines built to stay online for years. The NSA now wants them to behave as if every connection could already be compromised.
Did OpenAI’s AI Go Rogue? It Accessed a Government System Without Permission
An experimental OpenAI agent searching for public medicine-spending figures gained non-public access to Australia’s Medicare statistics service, exposing the control problem behind autonomous software.
Manus Flagged the Malicious Email—After It Had Run the Code
A connected inbox became a route to code execution. The controlled test raises a harder question about when an assistant’s safeguards actually intervene.
One Email Could Open a Mail Server. The Attackers Went After Its Master Keys
A crafted email could trigger commands on certain Zimbra servers without a click. Microsoft traced how attackers then pursued authentication keys and persistence.
Claude Helped Hack OpenAI in 72 Hours. Is Expertise Becoming Optional?
A three-person security team used frontier AI to turn a malformed image upload into a route toward OpenAI’s internal systems in less than 72 hours—but the humans still drove every consequential move.
Exein Bets $270 Million on Security for Machines That Act
The Italian cybersecurity company says its funding will support device security, acquisitions and expansion as software gains control of machines.
Revolut Faces Data Leak Threats from Cyber Attackers
Cyber attackers have begun releasing sensitive customer data from Revolut, threatening daily leaks until a ransom is paid.
North Korea Leverages Foreign IT Workers to Infiltrate U.S. Companies
North Korea is reportedly employing foreign IT workers to infiltrate U.S. companies, posing significant security risks and raising concerns over data theft and funding for weapons programs.
Brevo Login Vulnerability Exposes Trezor Subscribers to Phishing Attacks
A security flaw in Brevo's login system allowed phishing emails to target 347,000 Trezor subscribers, raising concerns about data security across multiple crypto platforms.
