Know what changed.
Trace why it matters.

Booz Allen tested frontier AI models against an isolated industrial-control testbed in its Chantilly, Virginia lab and reported that they completed all eight attack objectives defined for the test. The reported activities included network mapping, associating equipment with known weaknesses, writing working exploits, pivoting into production-control systems, and taking administrative control of and moving a robotic arm.

The test demonstrated successful model actions across multiple stages of an OT attack scenario, including exploitation, lateral movement into production systems, and physical manipulation of equipment.

Two of 31 newly disclosed flaws in Progress Software’s MarkLogic Server could allow an account with some management privileges to take administrator control of the server. Progress published all 31 flaws on October 12 and rated five of them critical.

The issue makes the boundary between limited management privileges and full server administration the relevant review point, rather than treating existing management permission as equivalent to administrator control.

Barracuda research describes phishing emails designed to influence both human recipients and AI assistants that process messages. Examples include hidden instructions telling a summarizing AI to portray an email as legitimate or urgent, or to add a fake priority action to change a vendor’s payment details.

An assistant may mislead users about which email deserves attention or what action it requires, while assistants with additional permissions might attempt unauthorized actions or disclose sensitive information, depending on capabilities and safeguards.