- AI Fire
- Posts
- 🚨 40+ CEOs Unite Against AI Threats
In partnership with
40+ CEOs are joining forces as AI cyber risks move closer to critical U.S. infrastructure. JPMorgan’s Jamie Dimon is helping bring the group together, but the companies involved and what they’re preparing for remain largely under wraps.
What's on FIRE 🔥
IN PARTNERSHIP WITH HUBSPOT
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.
AI INSIGHTS
JPMorgan Chase CEO Jamie Dimon is personally recruiting leaders from 40+ companies to expand the Alliance for Critical Infrastructure (ACI) as AI-driven cyber risks become a bigger concern. Main points:
ACI already existed before this AI push and was originally built for cross-industry coordination during cyberattacks, physical attacks, geopolitical disruptions, and other systemic crises.
Dimon started the new outreach in July and personally contacted CEOs across banking, technology, energy, utilities, telecom, and transportation.
The identities of most of the 40+ companies and the number that have formally joined are still undisclosed.
ACI wants companies to share how they are deploying AI, what vulnerabilities appear, and what safeguards or joint responses may be needed.
Dimon is helping recruit companies, but he says he is only a “support cast member,” rather than the person formally running the alliance.
That is why ACI overlaps with the U.S. government’s GOLD EAGLE initiative, launched in July. GOLD EAGLE uses frontier AI to coordinate vulnerability discovery, verification, prioritization, and remediation across government and critical infrastructure.
PRESENTED BY MESH API
Know Exactly Who's Spending Your AI Budget.
Every AI request leaves a trail. Mesh gives engineering and finance complete visibility into who used which model, how many tokens were consumed and where your AI budget is going.
Stop guessing. Start governing AI spend.
Connect once, switch between GPT, Claude, Gemini and hundreds more whenever you want, while automatically routing requests for 40% lower costs and 99.99% AI response rate.
AI SOURCES FROM AI FIRE
1. FREE: Qwen 3.8 Max Is Here. I Tested If It Can Beat Today’s BEST Open Models. Qwen 3.8 Max has arrived, but the real question is whether Alibaba’s newest model can compete with the strongest AI systems available today.
2. Create Ultra Realistic AI Videos with Realistic Motion Voice (Feel Like They Were Shot in Real Life). You’ll see how to combine the right prompts, visuals, and audio techniques to create realistic characters, environments that feel like they came from a professional production.
3. 3 AI Video Generators That Are Actually Free With Daily Credits. I tested SnapGen AI, TikTok Symphony, and Google Flow to see how much you can create without paying. You’ll see which options give 0-credit models, weekly credits, or 50 free credits every day.
TODAY IN AI
AI HIGHLIGHTS
💬 Jack Dorsey’s Block just open-sourced Buzz, a free self-hosted workspace where humans and AI agents like Claude, Codex & Goose can chat, review code, run workflows, and work in the same rooms.
🧬 Scientists at Stanford & Arc Institute used Evo to design viruses from scratch. Researchers tested 285 AI-generated genomes and found 16 working viruses that could infect bacteria, a major milestone for AI biology.
🧠 Anthropic just improved Fable 5’s biology safeguards, cutting biology-related fallbacks by about 85%. Users should now get better help with everyday health, lab results, symptoms, education, and some clinical tasks.
🗺️ Google Maps added AI agents to Ask Maps. It can now order food, compare real-time hotel prices, find events & tickets, and use Gmail context for personalized suggestions. The new agent features are rolling out first in the U.S.
🚀 ByteDance is reportedly training an AI model with up to 10 trillion parameters, roughly 3× larger than Moonshot’s Kimi K3. The TikTok owner wants to build one of the world’s largest models and compete directly with Anthropic.
💰 Big AI Fundraising: Omilia raised $67 million to expand its AI customer support platform after growing annual recurring revenue 10x to $60 million and deploying voice ordering across 1,000+ Taco Bell locations.
HOT PAPERS OF THE WEEK
1/ JD brings real-time AI video editing to 30 FPS
JoyAI-Video-Edit from JD’s Joy Future Academy introduces a 16B autoregressive diffusion model for open-ended video editing. It can edit 720p video at around 30 FPS on a single NVIDIA B200 GPU while keeping subjects, motion, and backgrounds consistent over long streams. Big shift: AI video editing is moving from offline generation toward live broadcasting and interactive content.
2/ Alibaba helps AI agents stay reliable on very long tasks
LongHorizon-Harness from Alibaba Group’s DreamX Team separates planning, execution, and verification through a Manage-Execute-Audit loop. It works with systems like Claude Code, OpenAI Codex, Claude Opus, and Qwen while keeping only verified task state between steps. Key result: Qwen 3.7-Plus jumps from 51.8% to 80.7% on WeaveBench.
3/ One image can now produce a more stable 3D world
InfiniSplat from Zhejiang University and Udeer.ai turns a single photo into a renderable 3D Gaussian Splatting scene. Instead of placing 3D points directly on the image grid, it aligns them with estimated scene surfaces for stronger geometry. Big impact: single-image 3D reconstruction could become more useful for AR/VR, spatial photos, and large viewpoint changes.
NEW EMPOWERED AI TOOLS
🛡️ Prelint reviews AI-written code against your docs, ADRs, and past decisions to catch product drift before a pull request ships.
🔌 Toolport is a free local MCP gateway that lets Claude, Cursor, VS Code, Codex, and 29+ agents share the same tools with up to 91% fewer tokens.
📊 Basedash Subscriptions sends fresh dashboard or chart snapshots to email or Slack on a daily, weekly, or monthly schedule.
🧠 Hexis gives companies one Git-backed place to manage AI skills, tools, and context with versioning, access controls, and MCP support.
AI BREAKTHROUGH
Researchers tested whether AI agents can handle open-ended AI research, giving them 6 days and a $3,000 budget to produce 2 research papers. Both papers were rejected.
The main runs used Claude Opus 4.8 with extra-high reasoning on OpenClaw, after dry runs across OpenAI and Anthropic models. An early GPT-5.3 Codex pilot could not reliably handle the scaffold. The agents were strong at execution:
Ran hundreds of experiments.
Debugged crashing GPU pods.
Produced camera-ready LaTeX without human intervention.
Dropped stronger marketing claims when experiments did not support them, rather than gaming the results.
The major weakness was research judgment. Automated reviews repeatedly flagged problems, but the agents mostly responded by narrowing claims and adding caveats.
The researchers who had worked on the same questions for months judged the papers and found that the agents lacked the ability to recognize when a research direction needed a fundamentally different idea → Both runs finished with more than half of the $3,000 budget unused.
We read your emails, comments, and poll replies daily
How would you rate today’s newsletter?Your feedback helps us create the best newsletter possible |
Hit reply and say Hello – we'd love to hear from you!
Like what you're reading? Forward it to friends, and they can sign up here.
Cheers,
The AI Fire Team
Reply