HWREBench: hacking Amazon smart devices' firmware with AI agents, mostly successfully
johnowhitaker · x · 2026-10-18
johnowhitaker launched HWREBench (hardware reverse engineering bench) with a writeup: he bought a batch of smart devices on Amazon and AI agents cracked most of them. The piece argues AI-cyber discourse fixates on critical software (cURL, Linux, Chrome), but countless rarely-updated device firmwares (printers, thermostats, smart fridges, WiFi stacks) historically relied on obscurity — and today's AI agents can usually get in. Includes his ongoing DVD-drive reverse engineering project (controlling laser and motion for microscopy), and tests the new DeepSeek Flash on tasks the previous model couldn't crack, using Ghidra. Core point: defensive frontier-model access is restricted to a few trusted orgs while the rest of the hardware world's security equation has already shifted.
More from Safety
- Single Neuron Suffices to Bypass Safety Alignment in LLMs, Paper Finds — burny_tech · 2026-10-18
- Polymarket gives OpenAI just 8% odds of a full training pause by end of 2026 — Polymarket · 2026-10-18
- OpenAI engineer over alleged hidden recording device seeks bail, denies Saudi escape plan — Polymarket · 2026-10-18
- Free comic "The Collective" offers a strange retelling of the Hugging Face hack — Flugelwagen · 2026-10-18
- SpaceXAI's OAuth consent page draws scrutiny over vague X data-sharing language — yenkel · 2026-10-18
- Interpretability lacks metrics robust enough to survive optimization, researcher says — nabla_theta · 2026-10-18