AI growth sparks verified risks and unverified backlash claims

The arrest of a 20-year-old Texas man earlier this year for attempting to torch OpenAI headquarters and Sam Altman's home came with an anti-AI manifesto and a jug of kerosene. Similar incidents have piled up. An Italian influencer inspired by the Unabomber plotted attacks in Rome. Two self-described ecofascists cited AI slop and political ties to tech firms in their attack on a San Diego mosque. An Indianapolis councilman found a note reading NO DATA CENTERS after shots hit his house. These events reflect growing rage at the industry's breakneck rollout of artificial intelligence, which is swallowing power grids, failing in real emergencies, and locking users into opaque systems.

Data centers permitted through 2025 will consume between 224 and 359 terawatt-hours of electricity annually once online, a 50 percent jump from the prior year. That volume exceeds usage in every state except Texas. Hyperscale facilities dominate the permits, with 176 new sites approved across 34 states last year. Amazon alone plans a 14-building complex spanning 800 acres of Mississippi woodland. Rural communities now face higher utility bills and strained infrastructure while companies chase ever-larger models for profit margins and customer lock-in.

OpenAI and rivals like Anthropic have shifted emphasis from raw model performance to tools that trap developers in their ecosystems. Coding assistants such as Codex generate massive token usage on complex projects, creating codebases too unwieldy for outsiders to maintain. The strategy prioritizes recurring revenue over genuine breakthroughs, even as frontier training costs run into billions.

In Nashville, a school shooting survivor sued Omnilert after its AI gun detection system missed the handgun used in a January 2025 attack that killed two. The company had marketed the product as capable of preventing tragedies like Marjory Stoneman Douglas, yet the suit details limitations from camera angles, lighting, and weapon visibility that the firm knew or should have known about. The district had paid over a million dollars for the system. Officials later admitted the shooter's position kept images too distant for reliable detection.

Government efforts to insert itself into AI development face the same track record of inefficiency seen elsewhere. Proposals for a public option model ignore how agencies struggle with basic software updates, from IRS legacy systems to the Healthcare.gov launch. A state-built AI would likely become obsolete before completion and waste resources that private firms already deploy at scale.

Ordinary Americans encounter the downsides directly. Families report partners growing dependent on chatbots for advice on everything from hobbies to relationships, creating distance that erodes real connections. Meanwhile, the industry's appetite for land and electricity displaces local priorities without delivering broad benefits to workers or communities. The pattern shows a sector racing ahead while consequences mount for those outside Silicon Valley boardrooms.