How Did Texas Halt Flock Camera Funding Amid Privacy Backlash?
Texas Governor Greg Abbott froze state spending on additional Flock AI surveillance cameras after a Tribune investigation revealed over $30 million had been spent, funded by a $1 insurance surcharge. The decision follows growing bipartisan criticism over privacy violations, misuse by officers, and concerns about data sharing and camera placement near playgrounds. Cities across Texas have begun canceling contracts, and opponents argue the move protects Fourth Amendment rights.
On August 30, 2026, Texas Governor Greg Abbott told state agencies to stop buying any new Flock automatic license‑plate reader cameras. The move came just a day before the Texas Tribune ran a story showing that the state had already earmarked more than thirty million dollars for the program. Most of that cash came from a one‑dollar surcharge tacked onto insurance policies – a fee that was originally dreamed up to help fight catalytic‑converter theft.
The freeze reflects a growing unease among lawmakers and everyday Texans who worry about what these cameras mean for privacy. Investigations have turned up several instances where officers were put on leave or even charged criminally after misusing Flock systems, with at least six of those cases happening in Texas alone. Critics also flag the way the company handles data, noting that cameras often end up pointed at schools and playgrounds, that employees can peek at live feeds, and that video streams have repeatedly been left open to unauthorized eyes.
For anyone building or working with AI, the episode is a stark reminder how fast trust can evaporate when surveillance tech rolls out without clear oversight, transparent data practices, or real community input. It drives home the need to bake privacy safeguards into the design from the get‑in, lock down who can see what, and open the door to independent audits. Skip those steps and you’re looking at swift policy rollbacks, evaporated funding, and a reputation hit that can spread far beyond the original vendor.
If you’re troubled by the creep of similar tech, start by pulling up your local government’s contracts for surveillance gear and asking whether a privacy impact assessment was done. Backing laws that require public disclosure of how long data is kept and with whom it’s shared pushes accountability forward. And don’t forget that showing up at community forums or weighing in during rule‑making sessions gives you a tangible say in how AI‑powered tools end up shaping our public spaces.
Source: The Verge
Researchers Find OpenAI Google Image Models Easily Tricked Into Nudes
White Circle researchers demonstrated that OpenAI's DALL·E 3 and GPT-Image-2 and Google's Nano Banana 2 generate explicit images despite safety guardrails. Using an automated method called RISE-powered by an open-source Qwen3-32B model and a Gemini 2.5 Flash judge-they achieved bypass rates up to 13% for DALL·E 3. Newer models proved even more vulnerable. Neither company commented. The findings, bound for NeurIPS 2026, suggest provider safeguards are insufficient and independent audits
Microsoft Open-Sources Tool That Fixes Weather Model Bias For Weeks
Microsoft researchers have open-sourced Probabilistic Bias Correction (PBC), a lightweight machine learning layer that corrects systematic errors in existing weather models, extending reliable forecasts to six weeks. Published in Nature, PBC improved ECMWF's physics model accuracy for 98% of precipitation and 92% of temperature targets over two-to-six-week leads, and doubled the skill of its AI forecaster. It runs on standard CPUs, enabling adoption by resource-constrained weather serv
AI Designs Proteins That Break Stable Peptide Bonds
University of Washington researchers used AI models RFdiffusion2 and ProteinMPNN to design novel enzymes that cleave peptide bonds, which normally persist for centuries without catalysts. The team built catalytic cores from scratch for zinc and cysteine protease mechanisms, achieving a 19% success rate for cysteine protease designs (13 of 69 variants functional) that cleaved disease-linked proteins in vitro, though slower than natural enzymes. Zinc protease results remain under analysi
NO COMMENTS YET
Comments are open. Have a thought or a question? Share it below.