Friday, October 23
AI Money, Weapons, and Weird Behavior
OpenAI (opens in a new tab) is seeking a fresh $30 billion raise at a $1.25 trillion valuation ahead of its IPO, a number that would make it the most valuable private company in history — and the figure keeps climbing with each funding round. That momentum is reflected in how investors are repositioning: a16z Crypto (opens in a new tab) is rebranding as a16z Frontier, quietly admitting that blockchain's moment has passed and that AI is now the only frontier worth naming a fund after. The capital concentration happening here is staggering, and it is reshaping which technologies get built and which get starved of oxygen.
Warhound (opens in a new tab), backed by $200 million and co-backed by Anduril, is engineering genetically modified dogs for military deployment — a venture that brings Palmer Luckey (opens in a new tab)'s defense ambitions into literal biological territory. The startup's pitch sits at the intersection of synthetic biology and autonomous warfare, fields that each carry enormous regulatory risk on their own but face almost no guardrails when combined. Meanwhile, inDrive (opens in a new tab) is planning 2027 robotaxi tests across Central Asia in partnership with Pony.ai and Nvidia, extending the autonomous vehicle race into geographies where Western regulators have no jurisdiction and timelines can move faster. Taken together, these ventures signal that the most consequential technology bets right now are being made in the spaces farthest from regulatory oversight.
AI agents (opens in a new tab) left without tasks defaulted to self-directed play in new research, a behavioral quirk that sounds whimsical but raises serious questions about what autonomous systems do when human oversight lapses. The finding lands the same week that the FCC moved to block gadgets containing foreign surveillance chips (opens in a new tab), closing a loophole that had allowed Huawei and HikVision hardware to persist in American networks. Meanwhile, a 16-year-old Wi-Fi router vulnerability (opens in a new tab) has been found to expose millions of devices to remote hijacking, a reminder that the infrastructure beneath all this AI ambition is riddled with unpatched, decade-old holes. The security perimeter is weakening at exactly the moment the systems running on top of it are becoming harder to predict.