Autonomous Human Weapons
AI Watches Every Meeting, Autonomous Weapons Bill Advances, and the 2029 Encryption Deadline
One in three American workers now report that AI bots have recorded their workplace meetings, according to a new survey. The figure reflects how rapidly AI-powered recording and transcription has moved from novelty to near-ubiquitous presence in roughly two years — often arriving without explicit employee consent, embedded in calendar invites. The implications extend beyond privacy: when every meeting is recorded, analyzed, and potentially indexed against performance reviews, the dynamics of workplace communication change fundamentally, replacing candor with performance.
The legal framework governing AI meeting surveillance is almost entirely absent. Most states still regulate workplace recording under consent laws written for audio recordings in the early 2000s. Whether clicking 'join meeting' on a platform displaying a recording notice constitutes informed consent for AI analysis, transcription storage, and potential use in employment decisions remains genuinely unsettled law.
A bipartisan bill — the Human Authority over Autonomous Weapons Act — is moving through Congress and would codify into federal law the existing Pentagon policy requiring human approval for any lethal strike. The measure has both Republican and Democratic co-sponsors, reflecting shared unease about autonomous weapons systems — drones, missiles, and ground robots — that are now genuinely capable of engaging targets without human confirmation. Defense technology advocates counter that requiring human approval creates a speed asymmetry against adversaries whose systems can engage in milliseconds; the bill's proponents respond that the decision to kill cannot be delegated to an algorithm regardless of tactical efficiency.
A Google executive is warning that AI-driven cybersecurity threats require organizations to transition to quantum-safe encryption standards by 2029. The concern is that quantum computers, accelerated by AI-optimized development, will be capable of breaking current RSA and elliptic curve encryption within that timeframe. If critical infrastructure, financial systems, and military communications are not migrated to post-quantum cryptography before that window closes, the vulnerability exposure would be enormous.
The AI chip market produced its own significant story: Etched, a startup building chips dedicated specifically to transformer inference rather than general-purpose GPU computation, is seeking a twenty-billion-dollar valuation, according to the Wall Street Journal. The company's architectural bet is that transformer models will dominate AI for the foreseeable future, and that purpose-built silicon can deliver radically better efficiency for running large language models than Nvidia's general-purpose approach. A twenty-billion-dollar valuation without significant revenue is a direct wager on that architectural permanence.