For business owners
AI is now testing the foundations
November 7, 2026 in summary: OpenAI closed the first $11B of its funding round at a $1.25 trillion valuation, launched a benchmark that tests AI agents on raw wet-lab data, and Google DeepMind's agents found weaknesses in 1024-bit RSA encryption. The models are now probing the foundations of science, security, and their own economics.
- Daily brief
- AI news
- Automation
November 7 was the day the AI story turned inward. The announcements were not about a cleverer chatbot or a flashier demo. They were about the foundations underneath everything else: the money flowing into the biggest AI company, the way AI is tested on real scientific work, and the encryption that protects the systems we all rely on. In each case, the models are now being measured against, and used to probe, the ground they stand on. The question running through all three is the same one a business owner should be asking: how solid is the ground, and who controls it?
The money is concentrating
OpenAI closed the first $11B of its latest funding round at a $1.25 trillion valuation, making it the most valuable startup ever, according to TechBuzz. The round started at $30B and was cut back as investors pulled back, which is the detail worth sitting with. Even the most valuable private company in history could not raise everything it wanted, on the terms it wanted. The capital is still flowing, but it is flowing with more caution than it was a year ago, and it is flowing toward one company above all others. That concentration matters for anyone building on AI. When a single supplier holds that much of the market's value, its pricing, its roadmap, and its outages become your business risk, not just its own.
The lab is the new test
OpenAI also launched BioBasicsBench, a 100-task benchmark that tests whether AI agents can read raw lab evidence like gel images and microscope data, according to Unite.AI. Its GPT-6 Astra scored 66.7 percent, ahead of GPT-5.6 Sol at 43.8 percent. The shift here is what the test measures. Benchmarks used to ask models to answer questions about science. This one asks them to do science, reading the messy, unlabelled evidence a researcher actually works with. The gap between the two generations is large, which tells you the capability is moving fast, but the more important point is the direction. AI is being graded on real work now, not on quiz questions, and the grades are being published. If you buy AI tools, that is the standard to hold vendors to: show the work on real tasks, not on a demo.
The locks are being stress tested
Google DeepMind said its AI agents found two weaknesses in old 1024-bit RSA encryption, including a signature forgery method up to 1,600x faster than before, according to Crypto Briefing. Modern 2048-bit keys are not affected. This is the second day in a row that AI has turned up flaws in something foundational, after OpenAI's security model found real vulnerabilities in Firefox and the Linux kernel yesterday. The pattern is now clear enough to name: the same models that write code and answer questions are being pointed at the locks, and they are finding the weak ones. This is not a crisis for anyone running modern systems, since 2048-bit keys hold up, but it is a warning for anything still running on legacy cryptography. Old encryption does not get safer with age, and the tools that can break it are getting cheaper and faster.
What this means for a small business
Three practical takeaways. First, check your own foundations. If any of your systems, vendors, or certificates still rely on 1024-bit RSA, that is now a known weak spot with a published, AI-accelerated attack. Modern defaults are fine, but legacy setups deserve a look. Second, change how you judge AI vendors. A model that can read raw lab data and score 66.7 percent on real tasks is a different kind of tool from one that chats well, and the vendors worth paying for are the ones who will show you results on work like yours. Third, plan around concentration. A $1.25 trillion valuation on a single AI supplier means the ecosystem depends on one company's health. Where you can, keep a second option warm, because the round being cut back from $30B is a reminder that even giants do not always get what they ask for.