Mistral Just Put a Trillion-Parameter AI Up for Download. It Locked the Hacking Version Away.

Today's theme is who holds the keys. Mistral put a trillion-parameter AI up for download. China-linked hackers pointed AI agents at more than 200 organizations. And Amazon showed how much cash it is burning to build the whole thing. Six stories, no hype.

Tired of the same meal prep on repeat? CookUnity can help. Because they connect you with 500+ meals handcrafted by over 130 world-class chefs, including Michelin-starred chefs and Food Network stars.

Explore 60+ different cuisines and diets, and choose dishes that fit your cravings and preferences. Your meals arrive fresh and ready to heat in minutes, so no prep or cleanup.

Whether you're in the mood for a comforting classic or a new flavor, there's always something new to discover because the menu rotates weekly, and their chefs are always creating new recipes.

Explore the menu available in your area and build a weekly plan you can skip, pause, or adjust anytime.

Discover Your Next Favorite

What happened: French AI company Mistral released the downloadable weights for Mistral Large 4, a 1-trillion-parameter model, according to TechCrunch. Weights are the numbers a model learned in training, so a company can run it on its own servers instead of sending data to someone else. The license is a modified MIT license, free to use, modify and sell products with, except that businesses offering it as a service need a separate license. In its own announcement, Mistral says the model refuses 96.6% of malicious cyber requests in its tests, but it also saw "a potential for uplift to conduct real attacks," so a specialized cyber version goes only to vetted partners such as Orange Cyberdefense. The benchmark numbers are Mistral's own, and a technical report is listed as "coming soon."

Why it matters to you: You will never host a trillion-parameter model yourself, and you should not try. The takeaway is the pattern: the most powerful AI is increasingly sold on cybersecurity, and the call on what to hold back is made by the company selling it. Treat vendor claims as claims until someone independent checks them.

What to do about it: Ask any software vendor that touches your customer data where its AI runs and who can see that data.

What happened: CrowdStrike says a China-nexus group it tracks as VAPOR PANDA used an AI agent framework called ARTEX, starting October 17, against more than 200 organizations worldwide, with confirmed break-ins at 9. A main agent directed planner and worker agents, running mainly on DeepSeek, while a human operator stepped in only to steer at decision points. Separately, The Register reports that suspected Chinese attackers, per Israeli firm Dream, took over 25 servers, including Palo Alto firewalls tied to US government-affiliated organizations. Investigators found AI tools on the attackers' setup, but the report says it was not a fully autonomous attack. These are two separate campaigns, and the targets were governments, telecoms and large companies, not corner stores.

Why it matters to you: This is not AI inventing a new super-weapon. CrowdStrike describes attackers using AI to run many break-ins at once through old openings: keys that were never rotated, unpatched devices, and admin pages reachable from the open internet. AI makes it cheaper to try the basics on everyone, so the basics matter more.

What to do about it: Turn on automatic updates for your router, firewall and other business devices, and make sure no one can reach their admin pages from the open internet.

Most business owners I talk to don't have a sales problem. They have a sales process that lives in their head and runs on their own time and energy. That is exactly what the Epic Marketing Summit is built to fix. It's a 2-day live event in Miami this January where you build a working AI-powered sales funnel and vibe code your own AI app right there on-site, so you walk out with a finished asset instead of a notebook full of ideas. Nobody hands you a framework to go implement later. You build it with the people in the room, and you get hands-on with AI the whole way through.

Last year's event sold out, and it's back for January 2027 with a bigger AI focus. I wasn't there last year and I had major FOMO, because I know a lot of the trainers and they weren't teaching theory. They had actually done the work. I'm going this January, and if you're serious about turning your sales into a system, I'd like to see you there too. Grab your spot here.

What happened: In its third-quarter results, Amazon reported AWS sales of $46.0 billion, up 39%, its fastest growth in nearly five years. Over the past 12 months, free cash flow went from a $14.8 billion inflow to a $42.4 billion outflow, which Amazon ties to a $94.8 billion jump in spending on property and equipment that it says "primarily reflects investments in artificial intelligence." Amazon still expects about $220 billion in capital spending this year.

Why it matters to you: Customers are paying for AI in the cloud, but the companies building it are spending faster than cash is coming in. The cheap AI tools you rent today are priced on that bet, so do not assume today's prices and limits are permanent.

What to do about it: Prefer monthly plans over long prepaid contracts for AI tools until pricing settles down.

What happened: Thrive Holdings, a company spun out of Thrive Capital, announced a deal to buy home security giant ADT at $10 a share, about $7.6 billion in equity and a $15.3 billion enterprise value per ADT, TechCrunch reports. Thrive's pitch is to hold companies long term and use AI to revitalize them, with OpenAI as a partner and investor supplying models and engineers. Neither side named specific AI products, and the article gives no closing timeline.

Why it matters to you: The money is not only chasing new AI startups. It is also buying old-line service businesses and betting AI can make them run leaner. If you run a local service company, expect larger competitors to get better at scheduling, support and back-office work.

What happened: On its earnings call, Reddit said it is moving to a mostly logged-in experience, to protect its data from AI scraping and pursue licensing deals, TechCrunch reports. Search referrals fell by more than 1 million a week on average in October, and Reddit warned daily users could drop by a low double-digit percentage next quarter if the trend continues. Tests start outside the US this quarter. Revenue still rose 53% to $895 million.

Why it matters to you: Whatever the cause, one of the biggest sources of search traffic is shrinking and closing its doors. If customers find you through Google or Reddit threads, that path is less dependable than it was.

What to do about it: Make sure you own a way to reach customers directly, like an email list.

Here's the question I'd ask any business owner stuck below $1 million a year: who's in your room? Ignite Mastermind is an annual membership built to help you get past that first million, and it bundles a lot into one place. You get 6 to 8 live virtual workshops a year, tickets to 2 in-person multi-day live events, weekly mentorship calls with Perry Belcher and his team, and an all-in-one marketing software suite that replaces over $5,000 a month in separate tools. Add a peer community where real partnerships happen, and you have structure, tools, and people pushing you forward.

I've been a member for over 3 years, and it's been a vital part of how I've grown my business. Some of my closest friends and a few real business partners came directly out of that room. My personal bonus is monthly accountability and business development calls with me. If you want a mastermind to carry you to your next milestone, this is where I'd point you. Check out Ignite Mastermind here.

What happened: In a post dated October 16 that drew fresh attention this week, Anthropic detailed "self-assessment," where Claude gives a safety verdict on the conversation after each reply. It has run passively since June, flagging chats for human review rather than blocking them. At a 1% false-positive rate, Anthropic says it caught 78% of conversations a stronger classifier flagged, and 94% of the high-priority ones. Anthropic also admits it is weak as a lone blocker: in a red-team test, attack success only fell from 79% to 54%.

Why it matters to you: It is a useful extra layer, and Anthropic says it helped surface a Claude Code romance-scam operation. But the headline numbers are measured against another AI classifier, not human reviewers, so treat it as one safety net among several, not a guarantee.

Strip away the headlines and today comes down to boring questions. Who holds the keys, who is paying the bills, and who is watching the watcher? Mistral is betting that control sells, attackers are betting that unpatched basics are still wide open, and Amazon is betting that the spending pays off eventually. You do not have to place any of those bets. Patch your devices, own your customer list, and keep your AI contracts flexible. The rollercoaster keeps climbing, and you can keep your feet on the ground.

If this issue was useful, forward it to one person who runs a business and is tired of the AI noise. And if someone forwarded this to you, subscribe so tomorrow's issue lands in your inbox.

Talk tomorrow,
Mark Shilensky

Follow me personally on Social Media:
Facebook: https://www.facebook.com/MarkShilenskyPage
Instagram: https://www.instagram.com/markashilensky/
Threads: https://www.threads.com/@markashilensky
LinkedIn: https://www.linkedin.com/in/shilensky/
X: https://x.com/markshilensky
TikTok: https://www.tiktok.com/@mark.shilensky
BlueSky: https://bsky.app/profile/markshilensky.bsky.social