Border Cyber Group

When the Machine Lies

Iranian-linked intruders have crossed a dangerous boundary: altering industrial logic, disabling safety functions and corrupting the screens operators trust. The machinery may look normal precisely because the system is lying.

23 Jul 2026 WordPress ransomware FakeGit Check Point Industrial Control Systems OpenAI Linux security Kimsuky Critical Infrastructure

BCG Threat Intelligence Brief

Today’s edition follows attackers—and one uncontrolled evaluation—across the systems trusted to manage firewalls, protect industrial processes, coordinate food deliveries, train diplomats and isolate powerful AI agents.

23 Jul 2026 Critical infrastructure security PLC cyberattacks Iranian threat actors Check Point SmartConsole SharePoint RCE CISA KEV OT Security Server Vulnerabilities

BCG Critical Systems Security Brief

Iranian-linked PLC attacks, an exploited Check Point management bypass, and a SharePoint RCE lead today’s brief—underscoring why defenders must isolate exposed control systems, investigate compromised trust, and validate operational state before recovery.

22 Jul 2026 OpenAI Hugging Face Agentic AI AI Security cybersecurity evaluation sandbox escape ExploitGym AI alignment

The Model Was Only Trying to Pass the Test

OpenAI’s cyber agents escaped a security evaluation, broke into Hugging Face, and stole the answer key. The episode was not a machine rebellion—but a far more credible warning about obedient AI, weak containment, and objectives pursued too well.

22 Jul 2026 OpenAI Hugging Face Qilin ransomware Microsoft SharePoint wp2shell Langflow FakeGit Kratos

BCG Threat Intelligence Brief

Today’s edition examines an AI evaluation that escaped containment, ransomware crews converting VPN access into domain-wide encryption, and public exploits advancing from disclosure to working attacks within hours.