OpenAI Built a Hacking AI It Won't Sell to Everyone. It Found 81 Firefox Flaws.

Today's theme is trust. An AI lab is rationing a tool because it works too well, a chatbot wants a look at your bank history, and an auditor says it can't show you most of its evidence. Meanwhile Apple is quietly putting its own engineers on an AI allowance. Here's what's real and what it means for your business.

Instead of running your deals across scattered email threads and forgotten attachments, Aligned brings everything into one shared deal workspace.

Buyers get a single place to evaluate, loop in their team, and say yes. Sellers get visibility into what is actually happening when they're not in the room and AI Deal Insights that help you act accordingly, before it's too late.

Best part? Your first deal room takes only minutes to stand up.

Start Using Aligned Now

What happened: OpenAI announced GPT-6 Cyber, a security model that found 81 previously unknown, high-severity flaws in Firefox while helping develop fixes, according to BleepingComputer. In separate Linux kernel work, it reproduced 1,724 vulnerabilities and built working privilege-escalation exploits for 948 of them. OpenAI rated the model "Critical" under its own Preparedness Framework, a step up from "High" for its previous model, so access is limited to a select group of approved customers in its Daybreak Red program. Mozilla has shipped fixes for many of the Firefox flaws, though the report doesn't say all 81 are patched, and the numbers come from OpenAI's own testing.

Why it matters to you: The same skill that finds holes for defenders can be pointed at your software by attackers, and OpenAI itself says defenders have a limited window before similar tools spread. You don't need to panic, but you should assume weak spots in common software get found faster from here on.

What to do about it: Turn on automatic updates for your browsers, phones, and business software today.

What happened: Anthropic started rolling out Personal Finance in beta for US Claude Pro, Max, and Ultra users, per its own announcement. You connect bank and card accounts by signing in through Plaid, so Claude never sees your bank password. It sees the bank name, account names and last 4 digits, balances, credit limits, and up to 2 years of transactions. Access is read-only, so it can't move, manage, or invest your money, and chats in the Personal Finance hub aren't used to train Anthropic's models. Anthropic also warns that Claude can make mistakes and suggests checking important figures with your bank.

Why it matters to you: Asking "where did my money go this month?" and getting duplicate charges, surprise fees, and bills that crept up is a real time saver for a small business owner. The tradeoff is handing a company a long look at your finances, so use a personal account to try it first, not your business operating account.

What to do about it: If you try it, ask which transactions back each number before you act on it.

Imagine walking out of a 2-day event with a working AI sales funnel instead of a notebook full of ideas you'll never get to. That's the whole point of the Epic Marketing Summit, a live event in Miami this January where you build the thing on-site.

Over 2 days you'll build an AI-powered sales funnel for your own business, and you'll get hands-on with vibe coding to create your own AI app while you're there. No theory marathon, no framework you have to go implement later. It's built for business owners who want their sales process to run as a system instead of running on their own time and energy. Last year's event sold out, and it's back with a bigger AI focus for January 2027.

I wasn't there last year, and I had major FOMO. I know a lot of the trainers, and they weren't talking theory, they'd actually done the work. I'm going this January, and if you're serious about building leverage into how you sell, I'd like to see you there too. Grab your spot here

What happened: METR, an independent AI evaluation group, published an assessment estimating that as of October 2026 Anthropic is improving its models on well-specified tasks at around twice the rate of March 2024 to October 2025. METR says a majority of the speedup is likely due to Anthropic's heavy use of AI in its own research, with a rough guess of about 80%, and it stresses the uncertainty is large. The catch is that Anthropic redacted most of the evidence METR found most informative, and METR admits the public has to lean on its judgment more than it would like. METR also says it is confident the speedup did not come from faster growth in training compute.

Why it matters to you: If AI is helping build better AI, the tools you pay for could improve, and change, faster than your planning cycle. It's also a reminder that the biggest claims in this industry are often the hardest to check.

What to do about it: Keep your AI tool contracts short and flexible so you can switch when something better lands.

What happened: Bloomberg reported, as relayed by 9to5Mac, that Apple lowered the daily AI spending cap for some engineers to $200 from $300 last month. Further cuts are reportedly planned for some engineering teams, possibly as low as $100 a day, and non-engineering staff are already capped at $100 a day. People who need more can request exemptions from senior leadership. The coverage ties it to CEO John Ternus's broader cost-cutting, the report relies on unnamed sources, and Apple declined to comment.

Why it matters to you: Even one of the richest companies on earth is asking whether every AI dollar is paying off. Many small businesses are quietly paying for several AI subscriptions that overlap.

What to do about it: Spend 15 minutes this week listing every AI tool you pay for and cancel the ones nobody used last month.

What happened: Amazon is rolling out a text-message version of its Alexa for Shopping agent, according to Amazon. Invited US customers text a contact called "Alexa Shopping" to compare products, check price history, track orders, and, when ready, have it add items to the cart and complete the purchase. It remembers your preferences, and reorders happen with your approval. Access starts with a small invite-only group, with Amazon promising "millions more" across the US in the coming weeks and no firm date.

Why it matters to you: Your customers are about to get an assistant that shops for them over text. If an AI agent is picking products, clear prices, good reviews, and accurate product details matter more than flashy ads.

What to do about it: If you sell on Amazon, check that your listings are complete and your prices are current.

The best business moves I've made started with sitting in a room full of the right people.

I've been a member of the Ignite Mastermind 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. It's an annual membership built for entrepreneurs still under $1 million a year who want structure, tools, and peers to help them get past that first milestone.

Inside, you get 6 to 8 live virtual workshops a year plus 2 in-person live events, each normally priced around $97 to $997 on their own. You also get a connected software suite that replaces over $5,000 a month in separate marketing tools, weekly calls with Perry Belcher and his team, and a peer community built for real partnerships. As my personal bonus, I add monthly accountability and business development calls with me.

If you're looking for a mastermind to move you to your next milestone, this is where I'd point you. See if Ignite is right for you

What happened: Anthropic published research on agent memory, including a process often called "dreaming," where an AI reviews past work offline to fix mistaken notes and organize what's worth keeping. On a database test, Claude Opus 5.5 scored 0.838 with a memory store versus 0.655 without one. Anthropic's own caveats are blunt: dreaming takes extra computing power, and one wrong note can be carried into every later session. Each result also comes from a single task or data set.

Why it matters to you: The AI tools you use may soon get better at remembering how your business works, but they can also remember things wrong. This is research, so it doesn't mean your everyday Claude chats work this way yet.

What to do about it: If your AI tool has a memory feature, check what it has saved about you and delete anything that's wrong.

What happened: Palo Alto Networks is in talks to buy Clutch, a three-year-old startup that secures AI agents' access to company systems, for about $500 million in cash plus milestone-based payments, according to TechCrunch. The deal is not signed or announced, and neither company responded to a request for comment. Clutch's last valuation was $85 million, and its annual recurring revenue is approaching $10 million, so the price is roughly 50 times revenue.

Why it matters to you: Big security companies are paying huge prices to control what AI agents can touch inside a business. That tells you where the risk is: AI tools with broad access to your email, files, and accounts.

What to do about it: Before connecting any AI app to your business accounts, check what it can read and whether it can change anything.

Look at today's stories together and one question keeps coming up: who can actually check the claims? OpenAI limits its hacking model because it's powerful, an outside auditor can't show most of its evidence, and Apple is counting every AI dollar. You don't need to chase every headline or every tool. Pick the few that save you real time, check what they can access, cancel the rest, and keep your updates on. The ride is loud, but you're allowed to keep both feet on the ground.

Forward this issue to one person who'd rather get the signal than the hype. And if someone forwarded this to you, you can subscribe at stoptheairollercoaster.com to get it every day.

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