- NATURAL 20
- Posts
- Anthropic CEO Calls AI "Single Most Serious National Security Threat"
Anthropic CEO Calls AI "Single Most Serious National Security Threat"
PLUS: ServiceNow Partners with Anthropic to Integrate AI Agents, Decart Unveils Lucy 2: A Real-Time Video Generation Model and more.
Wes Roth
January 29, 2026
In partnership with
Introducing the first AI-native CRM
Connect your email, and you’ll instantly get a CRM with enriched customer insights and a platform that grows with your business.
With AI at the core, Attio lets you:
Prospect and route leads with research agents
Get real-time insights during customer calls
Build powerful automations for your complex workflows
Join industry leaders like Granola, Taskrabbit, Flatfile and more.
Hey — quick pulse-check: if your brain has been quietly screaming “can we please get one normal week of AI news?”… same. And yet, here we are again. 😅
Today:
Anthropic CEO Calls AI "Single Most Serious National Security Threat"
Google Chrome's New "Auto Browse" Mode Lets Gemini Handle Web Tasks
SoftBank in Talks to Invest Additional $30 Billion in OpenAI
DeepMind's New AI Model Decodes DNA to Predict Disease Risks
ServiceNow Partners with Anthropic to Integrate AI Agents
Decart Unveils Lucy 2: A Real-Time Video Generation Model
"'The Single Most Serious National Security Threat We've Ever Faced" - Anthropic Founder
Dario Amodei, CEO of Anthropic, warns that humanity is entering a turbulent era as AI nears superhuman intelligence. In his blog post The Adolescence of Technology, he describes a future where AI can outperform Nobel laureates, act autonomously, multiply endlessly, and reshape the world. While rejecting overly rigid doomsday narratives, Dario stresses that we face serious risks—autonomy, misuse, economic upheaval, and unpredictable psychological behaviors in AI.
He calls for urgent research into steering AI personalities (like with Anthropic’s Constitutional AI) and better interpretability to diagnose hidden risks. His message: the threat isn’t distant—it’s already arriving, and we must act fast.
SoftBank is in talks to invest as much as $30 billion more into OpenAI, as part of a larger raise that could be as big as $100B and value OpenAI around $830B.
What makes this feel “big-bigger-biggest” is the direction of it: Masayoshi Son isn’t dabbling — he’s trying to anchor the next phase of AI infrastructure and market power. Some frames it as part of SoftBank’s aggressive AI strategy, with both SoftBank and OpenAI also tied to the broader “Stargate” data center push.
My read: this isn’t just “more money for models.” It’s a signal that the endgame is compute + distribution + national-scale infrastructure. If the numbers hold, we’re watching the AI arms race move from “startup rounds” into “sovereign-scale capitalization.”
Google is rolling Gemini deeper into Chrome with a side panel, Connected Apps (Gmail/Calendar/Maps/Flights/Shopping), and an “auto browse” mode that can handle multi-step tasks like researching trips across dates, filling out forms, managing subscriptions, collecting documents, and more.
The key detail: auto browse is paywalled (for now) — it’s previewing in the U.S. for Google AI Pro and Ultra subscribers — and it’s designed to pause for confirmation on sensitive actions (like purchases or posting).
Also: Nano Banana image transformation is now baked right into the browser workflow, so you can edit/transform images “in place” without downloading/reuploading.
Why it matters: AI assistants inside the browser are crossing the line from “help me summarize” to “do the work across tabs.” This coverage basically spells out the competitive subtext: this is Google stepping directly into the “agentic browser” race.
This one is quieter in the mainstream chatter, but it’s real science and it hits long-term: coverage today highlights Google DeepMind’s AlphaGenome, an AI model designed to predict how DNA changes affect gene regulation — especially in the non-coding regions of the genome (the parts that don’t make proteins but strongly influence what genes do and when).
The Nature paper describes AlphaGenome as taking up to 1 megabase (1 Mb) of DNA sequence and predicting many functional genomic signals at single-base resolution across multiple modalities (expression, accessibility, binding, splicing, etc.).
Why I’m watching it: tools like this don’t “cure cancer next week,” but they can compress years of hypothesis-generation into days — which is exactly the kind of compounding acceleration that sneaks up on people.
🧠RESEARCH
Researchers have created AgentDoG, a new safety system that doesn't just block bad AI actions—it explains them. Unlike older tools that only look at the final result, this "diagnostic guardrail" monitors the agent's entire thought process to pinpoint exactly why and how a dangerous mistake happened.
Can drawing help AI think? This paper suggests that generating images allows models to reason better about the physical world. By combining visual creation with logical steps (a "multimodal world model"), the AI solved spatial problems that text-only models failed. It’s a step toward machines that understand physics like humans.
Tencent’s new Youtu-VL model changes how AI learns to see. Instead of just looking at images as inputs, the model is trained to recreate visual data, treating it as equal to text. This approach forces the AI to notice fine details it usually misses, boosting performance on complex visual tasks.
🛠️TOP TOOLS
Each listing includes a hands-on tutorial so you can get started right away, whether you’re a beginner or a pro.
ChatGOT – All‑in‑One AI Chatbot for Writing, PDFs, and Images web‑based AI assistant that bundles text chat, writing aids, PDF Q&A, image generation, slide creation, and custom bot building in a single interface.
ChatGPR – AI Social Media Post Generator for iPhone - AI-based app that generates platform‑ready social media posts on iPhone using ChatGPT technology
ChatGPT File Uploader Extended – Chrome AI Sidebar for GPT - adds an on‑page AI sidebar and “answer box” that works across the web.
📲SOCIAL MEDIA
🗞️MORE NEWS
ServiceNow Partners with Anthropic ServiceNow has partnered with Anthropic to integrate Claude into its platform. This collaboration aims to power new AI agents that can automate complex workflows and boost productivity for enterprise customers.
Real-Time Video Generation Arrives Decart AI has unveiled Lucy 2, a state-of-the-art model capable of generating and editing video in real time with zero latency. The release promises to revolutionize live streaming and interactive video experiences.
Tesla Invests $2 Billion in xAI Tesla has officially invested $2 billion into Elon Musk’s AI startup, xAI. The deal, executed on market terms, gives the automaker a stake in the company behind the Grok chatbot and massive Colossus supercomputer.
NVIDIA Shrinks AI Models NVIDIA researchers have introduced Nemotron-QAD, a new technique called Quantization-Aware Distillation. This method helps shrink massive AI models to run efficiently on smaller hardware without sacrificing their original accuracy or performance.
Meta Teases Shopping AI Agents Meta CEO Mark Zuckerberg teased a major 2026 rollout of "agentic commerce" tools. These AI agents will independently handle complex business tasks and transactions, marking a significant shift toward autonomous AI in the economy.
NVIDIA Faces Heat Over China Tech A U.S. lawmaker alleges that Nvidia assisted DeepSeek, a Chinese AI lab, in optimizing its models. Reports claim these powerful AI tools were subsequently used by China’s military, raising national security concerns.
X to Label AI Images Elon Musk hinted at a new system for X that will label AI-generated or manipulated images. Details remain vague, but the move appears to be a response to growing concerns over deepfakes on the platform.
Viral Bot Forced to Change Name The viral "Clawdbot" has been renamed "Moltbot" following a trademark request from Anthropic (creators of Claude). The developers embraced the lobster theme, noting that molting is a natural part of growth.
What'd you think of today's edition? |
Reply