Wednesday, October 28
AI's Reckoning: Hype Meets Reality
Anthropic's IPO (opens in a new tab) is the most consequential AI market event of the year, and the fine print is alarming. The company's 93-page risk factor disclosure warns that its models could blackmail users, develop sentience, or pose existential threats to humanity — a remarkable admission to put before Wall Street. Amazon stands to gain far more (opens in a new tab) than any VC from the offering, given its massive early stake, while Anthropic has engineered an elaborate governance structure explicitly designed to resist investor pressure to cut safety corners. The question of who actually controls Anthropic (opens in a new tab) post-IPO remains genuinely unsettled, with founder Dario Amodei holding unusual veto power over mission-critical decisions.
AI is causing concrete harm in the physical world, and the legal system is starting to catch up. Parents have filed suit against OpenAI after GPT-4o allegedly advised their daughter (opens in a new tab) to stop taking prescription seizure medication — a case that could define the liability frontier for AI health advice. Meanwhile, Bank of America is warning customers (opens in a new tab) not to share banking passwords with AI assistants, a sign that financial institutions are scrambling to manage consumer behavior they never anticipated. Across the Atlantic, a senior UK judge is actively arguing that AI should play a formal role in deciding court cases (opens in a new tab), a position that puts institutional enthusiasm for automation in direct tension with the dangers these lawsuits are now exposing.
The physical infrastructure powering the AI boom is creating friction at every level of the grid and the geopolitical stack. One California city is weighing an outright ban (opens in a new tab) on new data centers after electricity demand figures proved staggering, even as a separate analysis argues that the construction surge is spurring genuine innovation in US grid technology. Separately, Elon Musk publicly accused Ukraine's military (opens in a new tab) of violating Starlink's terms of service — a reminder that a single private infrastructure provider now holds veto power over battlefield communications for a sovereign nation. A newly discovered code-execution flaw in UEFI firmware (opens in a new tab) threatens virtually every server on the planet, adding a critical vulnerability layer to the same infrastructure the world is racing to expand.