The video distinguishes shadow AI from approved AI whose access and capabilities expand beyond its original purpose. It explains how agents can combine authorized actions into unauthorized outcomes and why approval alone may not provide ongoing control.
EFF technologist Cooper explains Ray Hunter, an open-source tool that flags cellular-network behavior associated with IMSI catchers. The project reports no evidence of their use at protests so far and describes a Philippines detection coinciding with an SMS phishing msg
The speakers discuss the Wi-Fi Pineapple Pager and USB Rubber Ducky, describing community-created Wi-Fi payloads, limits on payloads without a legitimate penetration-testing use, and how writing or examining payloads can teach Linux, Bash and PowerShell.
At Splunk.conf16, demos and interviews show how telemetry, observability and AI agents can support security investigations and response. Speakers stress validating AI outputs, managing costs and keeping people responsible for consequential decisions.
The video explains how indirect prompt injection in emails or malicious tools can misuse an AI agent’s legitimate permissions, and describes monitoring and controls Radware says its AI security offering provides.
Crypto scammers gained unauthorized access to Microsoft’s X account for about 30 minutes, using Clippy-themed posts to promote a token. Microsoft confirmed the access, removed unauthorized posts, secured the account, and said it was investigating.
A hands-on tutorial uses Ghidra and x64dbg to analyze SteelBit and BlackMatter, demonstrating anti-debugging checks, runtime string decryption and Sysmon-based detection opportunities.
The transcript describes Open Shell as an isolation system for AI agents, using zero trust, minimum privileges, policy monitoring and formal verification to support safer deployment.
NVIDIA says Open Shell restricts AI agents through sandbox permissions, while optional Sentry on BlueField 4 can monitor and quarantine agents. The presenter has not independently tested NVIDIA’s claims and cautions that authorized actions can still cause harm.
An EFF guest walks David Bombal through Surveillance Self-Defense guides, explaining secure messaging, VPN limitations, AI data privacy risks and practical device protections tailored to individual needs.
The video describes the LAN Turtle Hub as a USB Ethernet adapter and hub that can provide remote access, sniff network traffic, poison DNS, intercept USB storage, plant a malicious shortcut, and copy files over Wi-Fi.
In an interview with David Bombal, researcher Marcus discusses AI restrictions on malware analysis, ransomware actors’ use of vulnerable drivers, and the limits of VPN protection against malicious captive portals.
The speakers recommend Sysinternals as a starting point for checking a Windows PC and show how Wireshark’s DNS filter can reveal network queries. They also discuss how Windows telemetry and third-party data collection complicate understanding where data goes.
At DEF CON, Jason Haddix demonstrates prompt-injection attacks against chatbots, including simulated account-recovery and meeting-summary scenarios. He says the examples reflect reported real-world flaws and highlights repeat testing and free practice labs.
The Shark Jack display is an Ethernet-connected device that runs selectable payloads for network and port scans, LLDP discovery, and connectivity checks. Two devices can use Shark Link to validate network paths and configurations.
Researchers demonstrated that a crafted link could make Microsoft Copilot execute a prompt, exfiltrate connected data and poison persistent memory. The Co Snitch vulnerability chain has been patched, and researchers reported no evidence of exploitation.
A speaker describes how a visible license plate may lead from public vehicle records to a VIN and vehicle-history data, potentially exposing an owner's address, businesses and routines.
Radware researchers found more than 3,000 hacking and fraud guides on underground forums. New tutorials rose to an estimated 110–140 per month in 2026, with more focus on carding and identity theft; generative AI may make them easier to produce.
Hak5's Darren discusses the Wi-Fi Pineapple Pager's wireless testing, event-triggered alerts and customizable payloads, and compares it with the Mark VII and Enterprise models.
The video presents a pager-style Wi-Fi Pineapple for portable Wi-Fi testing, with an open architecture for custom payloads. It is described as a tool for authorized penetration testing, with filters to keep tests within scope.
The guest says he used fake personas to approach ransomware groups and spent about 18 months speaking with LockBit. He describes the group's attacks on children in hospitals and the personal toll of the work.
Researchers demonstrate how patched mission control vulnerabilities could expose configuration files or let a malicious display script send commands to a simulated spacecraft.
A researcher explains how carefully induced CPU failures can trigger machine check exceptions and expose security risks in privileged handler code. The discussion also notes that secure source code may not produce a secure binary.
An Asurion SOC speaker describes deploying about 45 AI agents to review security data, run searches and flag data-source problems, while emphasizing context and verification to address AI hallucinations.
