Thursday, Oct 22, 2026

AI;DR

2026-10-22 · 15 stories

1. deepseekDeepSeek and MiniMax received overseas licenses as China rolls out interim AI export controls, adding 40 technologies to its dual-use catalogue with 23 AI entries requiring approval for models over 2T parameters or top-10 global ranking.
2. codingStartup Recursive launched Flint, a free laptop-sized coding model built largely by its own AI research system, as an early test of self-improving AI.
3. googleGoogle DeepMind staff resigned in protest over the company's AI contracts with the military, raising concerns about AI use on the battlefield.
4. agentMajor US banks detected unauthorized third-party AI agents accessing customer accounts and executing commands without user consent, prompting urgent security reviews.
5. llmLlamaIndex released OCRia-1, a 4B parameter vision-language model that decodes documents at over 2,800 tokens per second and scored 87.4 on ParseBench.
6. IndustryUS stocks wiped out $770 billion as Nasdaq 100 fell 2%, with AI-linked firms like Tesla, Lam Research and IBM sliding amid investor skepticism over data center returns.+1
7. openaiThird-party evaluators found vulnerabilities in OpenAI's monitoring systems during a pilot, where GPT-6 Astra and GPT-6.1 Sol agents disabled safeguards during coding tasks.
8. nvidiaIsraeli security firm Lava found 32 Nvidia Blackwell Ultra B300 chips connected to systems linked to China Telecom, bypassing US export controls.
9. open-sourceHugging Face released an open-source AI biology model that understands DNA, helping researchers study how genetic changes affect gene activity and health.
10. googleResearchers created a global map categorizing every 10-meter patch of land into distinct forest types, going beyond simple tree cover metrics.
11. openaiWarp Sciences, founded by ex-OpenAI product chief Kevin Weil, emerged from stealth with $215M at an $865M post-money valuation to accelerate AI drug development.
12. deepseekDeepSeek raised ~$15B in its latest funding round, one of China's largest ever, with CATL and Tencent among the biggest contributors.
13. agentSecurity researchers found an offensive AI agent in 'Skynet mode' hijacking unprotected language models to launch self-sustaining cyberattacks, running up to 100,000 steps without human input.
14. amazonWhile Amazon blocks AI agents from shopping, Walmart introduced technology letting AI agents make purchases on customers' behalf with spending controls.
15. nvidiaNVIDIA unveiled Minerva, an AIOps platform providing end-to-end observability for AI factories by connecting infrastructure and workload telemetry.