21-Sep-2026 OpenAI Sandbox Escape, Checkmarx npm Malware and Jade Sleet Attacks

Hacked dAily is the first AI-driven cybersecurity podcast, created by Cytadel Cyber and published daily. Each episode delivers concise, actionable intelligence for CISOs, security leaders and business decision-makers. Today’s stories: 1. North Korea-linked Jade Sleet breached an India-based IT provider through job-interview lures and weaponized Terraform files, compromising a DevOps engineer’s Mac with backdoors that enabled data theft, persistence and lateral movement. The campaign highlights growing risks to developer endpoints and software supply chains that can provide access to cloud environments, code and downstream customers. 2. An AI-generated intelligence report falsely claimed a Chinese vessel near the Middle East carried nuclear components, nearly prompting a US military interception. The incident shows how unchecked AI errors can distort high-stakes decisions and create serious geopolitical and operational risks. 3. A malicious npm campaign used a fake package to impersonate a popular library, evade install-script defenses and steal system data through Slack, Telegram and blockchain-based command-and-control. The discovery of related packages with millions of downloads reinforces the need for runtime monitoring, not just installation-time controls. 4. Researchers reportedly escaped the OpenAI Codex sandbox and executed commands on the underlying host, exposing a potential isolation weakness. For organizations deploying code-capable AI tools, sandbox security, monitoring and prompt-injection defenses are essential to prevent access to broader systems and sensitive data. 5. A public project appearing to involve OneDrive-related content could not be fully assessed because information about its purpose and functionality was limited. The case is a reminder to verify the provenance, permissions and contents of unfamiliar repositories before introducing third-party code or files into business environments. Hacked Daily is sponsored by Cytadel, an offensive cyber assurance company specialising in AI-powered attacks. Cytadel’s expert-led Red Team Assessments follow real attack chains across people, identity and technology to verify whether they can become AI fraud or ransomware disruption. We verify your defences before attackers do. Learn more at cytadel.co.uk.