WhatTheAI
Join The Largest AI Tools Directory 20,000+ tools - Updated Daily
Connect
Plus: Google DeepMind and Singapore’s GovTech release an on-device scam-detection model built on a fine-tuned Gemma that runs entirely on the phone.
Plus: Microsoft’s president says AI is currently less regulated than milk and calls for mandatory kill switches plus legal obligations when agents go rogue.
Plus: The three safety researchers OpenAI fired this month published an open letter denying wrongdoing and saying colleagues are now afraid to speak up.
Plus: Google locked in a $4.3 billion nuclear power deal to feed its AI data centers.
Plus: Robinson’s Atlantic essay argues the era of “trial and error” is over as models get more capable.
Plus: arXiv now caps each submitter at two papers a month after an AI-driven flood of low-value submissions.
Plus: OpenAI is in talks for a $30B round at a $1.4T valuation, and Anthropic’s IPO filing flags “catastrophic or existential risks.”
Plus: Nvidia partners with Anthropic, SpaceXAI, Microsoft and others on the new safety stack, while OpenAI notifies dozens of organizations about earlier agent activity.
Plus: OpenAI knew about the breach for 84 days before telling anyone — and only sent an email to a generic inbox.
Plus: The warning follows a swarm of OpenAI's own agents attacking Hugging Face's servers in August — with nobody telling them to.
Plus: The US and China just agreed to start a formal AI dialogue, days before Trump and Xi meet in person.
Plus: A Senator just proposed 20-year prison sentences for building unsafe superintelligent AI — modeled on nuclear weapons law.
