AutoProber vs OpenAI Privacy Filter

AutoProber is an open-source hardware security research platform that puts an LLM agent in control of a physical CNC machine to autonomously probe circuit boards. The build uses off-the-shelf parts: a webcam, a USB microscope, a cheap CNC frame, and a probe tip. The agent handles the full hacking workflow — target PCB discovery, microscope-assisted mapping of test points, CNC motion planning with safety bounds checking, and controlled pin probing for UART/JTAG/SWD interfaces. The software stack is pure Python. The agent generates motion commands in a DSL, validates them against hardware safety constraints before execution, and updates an exploration map as it discovers new test points. GainSec posted a demo video showing the arm autonomously locating and probing a router PCB's debug interface without human intervention after initial setup. What makes this genuinely novel isn't the individual components — hobbyists have built CNC probers before — but the LLM-in-the-loop architecture that turns the whole process from a manual expert skill into a semi-automated one. Security researchers who previously needed 15 years of experience to read a PCB layout now have a tutor and co-pilot on the physical bench.

OpenAI released Privacy Filter on April 22, 2026 — a 1.5B-parameter open-weight model for detecting and redacting personally identifiable information from text before it ever reaches a cloud API. The model runs fully locally, handles 128,000 tokens in a single pass, and achieves a 96% F1 score across eight PII categories: names, addresses, emails, phone numbers, URLs, dates, account numbers, and secrets. Unlike traditional regex-based PII scrubbers that choke on unstructured text and context-dependent references, Privacy Filter uses a fine-tuned language model to understand semantic context — it catches "call me at the usual number" type references that pattern matchers miss entirely. The model ships with only 50M active parameters at inference time via sparse activation, keeping latency low enough for preprocessing pipelines. Available on Hugging Face and GitHub under Apache 2.0, Privacy Filter solves a real bottleneck: enterprises and regulated industries have been unable to safely pipe sensitive documents through LLMs at scale. OpenAI explicitly warns it should be treated as a "redaction aid, not a safety guarantee," which is unusually honest for a model card — and a sensible framing for high-stakes medical or legal workflows.