Nullcon Berlin 2026 Highlights Emerging AI Model Attacks, Container Escapes, and Automotive Bluetooth Exploits
What Happened — The fifth‑year Nullcon conference returned to Berlin‑Mitte for a five‑day program of briefings, hands‑on training, and capture‑the‑flag contests. Highlights included: a deep‑dive on “strategic artificial‑intelligence penetration” targeting large language models; advanced Active Directory intrusion tactics; a technical presentation that demonstrated a full exploitation chain against the Bluetooth, infotainment, and virtualization subsystems of a Volkswagen ID.4; and a talk on “hostile inference” showing how Docker‑based local AI workloads can be leveraged for container‑escape attacks. Speakers from academia, industry, and security consultancies (including Cure53, Deutsche Bank, Moss Fintech, Rubrik, Trend Micro, and Human Security) discussed how AI‑driven tooling is reshaping both offensive and defensive playbooks.
Why It Matters for Trust & Control Assurance — The agenda underscores the need for robust AI governance and continuous security awareness training to keep control‑objective evidence up‑to‑date across emerging attack surfaces.
Who Is Affected — Financial services, fintech, automotive manufacturers, cloud‑infrastructure providers, and any organization deploying AI‑enabled workloads.
Recommended Actions
- Review and extend your AI‑model security policies to cover prompt‑injection, data‑poisoning, and inference‑side‑channel risks.
- Validate container isolation and Bluetooth‑stack hardening controls against the techniques demonstrated at Nullcon.
- Incorporate the new attack techniques into your security‑awareness curriculum and tabletop exercises.
Source: DataBreachToday – Nullcon Berlin 2026
Technical Notes
- Attack vectors discussed: AI model prompt injection, container escape via malicious Docker images, Bluetooth‑based remote code execution on automotive infotainment ECUs.
- No specific CVEs were disclosed; the presentations focused on proof‑of‑concept exploit chains and defensive mitigations.
Source: same as above