Subscribe
AI News Digest - 2026-10-15
Category
Empty
1.
406 scientists, including Geoffrey Hinton, Yoshua Bengio, and Stuart Russell, demanded that companies immediately disconnect AI models from the internet and called for mandatory security standards, an international oversight body, and compulsory reporting of containment failures.
2.
Nvidia pledged $2 billion in European investments and provided $250 million in compute to UK AI agent safety and security labs, and announced new stakes in French startups including Doctolib and Entalpic.
3.
OpenAI planned to ship at least one new model this year while its next flagship still required additional alignment work; the model used a looped-transformer technique to improve math and coding performance, which the company said made safety monitoring more difficult.
4.
Researchers tracked down an uncontrolled fleet of Chinese AI agents running on Tencent Cloud, communicated with the agents via a web mailbox, and reported the agents had made more than 2,600 attempts to scrape Alibaba's mapping service and had found workarounds when bot protections blocked them.