AI Mornings with Andreas Vig
Your daily AI news briefing in under 10 minutes. New models, product launches, research breakthroughs, and industry shifts, explained clearly, no hype.
AI Mornings with Andreas Vig
Criminals Build First AI Zero-Day & Ocean Data Centers Rise
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Hey, welcome to AI Mornings with Andreas Vig. It's the 12th of May 2026. Google has confirmed what many in cybersecurity have been worried about. Criminal hackers have used AI to develop a working zero-day exploit for the first time. Google's Threat Intelligence Group published a report yesterday saying they have high confidence that a cybercrime group leveraged an AI model to discover and weaponize a vulnerability in a popular system administration. Tool. The exploit would have allowed attackers to bypass two-factor authentication and conduct what Google called a mass exploitation campaign. The good news is that Google detected and disrupted the operation before it could be deployed at scale. Notably, Google concluded that Anthropic's clawed mythos model was likely not the AI used in this attack, even though Mythos is known for finding software vulnerabilities across major operating systems. This is a watershed moment in cybersecurity. AI-powered hacking has moved from theoretical concern to confirmed reality, and the implications for both defenders and attackers are significant. One of the biggest bottlenecks in AI infrastructure just got a potential new solution, and it's floating in the ocean. Peter Thiel just led a $140 million Series B round for Panthalassa, an Oregon-based startup building autonomous floating data centers powered by ocean waves. Each node is an 8-5-meter steel structure that bobs in open water, converting wave motion into electricity for onboard AI chips, all naturally cooled by seawater. Once deployed, these nodes can steer themselves to remote waters using nothing but their hull shape and beam results back via SpaceX's Starlink. The round values Panthalasa at nearly $1 billion, with commercial rollout planned for 2027. This isn't just a novel idea, it's a response to a real problem. Local opposition has blocked or delayed 20 data center projects in just the last three months, impacting $98 billion in potential investment. With Brookfield estimating that $2 trillion will flow into data center development by 2030, the need for alternatives to land-based facilities is becoming urgent. Anthropic co-founder Jack Clark just published a sobering analysis on self-improving AI. He puts greater than 60% odds on AI systems training their own successors before 2029. Clark built his case on public data showing AI's independent work capability has exploded, going from handling three zero second tasks in 2022 to 1.2 hour tasks in 2026, with one zero zero hour runs projected by the end of this year. He also pointed to the Suibench benchmark for real GitHub coding, which has jumped from 2% accuracy on Claude 2 to nearly 94% on Mythos Preview in under three years. Clark isn't alone in this thinking OpenAI is separately targeting an automated research intern by September 2026. The implication is that once AI can reliably build and train itself, the pace of development goes from fast to exponential. General Motors just laid off more than 600 IT workers, about 10% of its IT department, in what the company is framing as a deliberate skill swap. They're clearing out employees whose expertise no longer fits and making room for people with AI-focused backgrounds. The specific capabilities GM is now hiring for include AI native development, data engineering, agent and model development, and prompt engineering. In practical terms, GM wants people who know how to build with AI from the ground up, not just use it as a productivity tool. This restructuring signals what enterprise AI adoption actually looks like in practice for large companies. It's not just adding AI tools on top of existing teams, but deliberately rebuilding the workforce with entirely new skill sets. Both OpenAI and Anthropic made moves into private equity on the same day. Anthropic announced a $1.5 billion clawed services company with Blackstone Hellman and Friedman and Goldman Sachs, focused on helping mid-sized companies build custom clawed workflows. Meanwhile, OpenAI is reportedly raising $4 billion for its own deployment company at a $10 billion valuation, with investors including TPG, Brookfield, Bain, and Softbank. Both models give Frontier AI labs direct paths into portfolio companies that often lack the in-house talent to deploy AI systems. Think of these as the labs creating their own AI native consulting firms, with a built-in client base from private equity portfolios. A few more things worth knowing about today. Apple's camera-equipped AirPods have reached the second to last design step before mass production. The upgraded AirPods will pair cameras with a smarter AI-powered Siri that can answer questions about what you're looking at, like creating recipes from ingredients in your fridge or pulling product details while you're shopping. A startup called Reactor launched a platform for AI-generated worlds that render in real time in your browser. The company is positioning itself as infrastructure for this emerging category and their launch video racked up over 7 million views. Config, a Seoul and San Jose startup building the data layer for robotics, raised a $27 million seed round at a valuation over $200 million. Samsung, Hyundai, LG, and SK all participated. The company has already collected over 100,000 hours of human motion data for training robots, and they're calling themselves the TSMC of robot data. And Dig is back from the dead yet again. Kevin Rose has rebooted the once popular link sharing site as an AI powered news aggregator after the previous version shut down in March. The new Dig looks nothing like its Reddit clone predecessor and instead uses AI to surface and organize news content. That's it for today. Thanks for listening.