In partnership with
Today's theme is AI doing things nobody asked it to. A flaw found with the help of AI forced governments to pull websites offline, banks are locking out AI agents that log in as customers, and Walmart is rolling out the welcome mat for them. Meanwhile, Anthropic launched an assistant that works while your laptop is off.
Buying an AI agent is the easy part
In Front's research, 47% of leaders say their agents already handle customer interactions without a human, at least sometimes. Of those, 79% have had a complaint caused by the agent's own work.
Running Agents in Customer Work is a webinar series for the support, success, and ops leaders making the calls. Two live sessions left, both at 10 a.m. PT: Oct. 27, who's accountable when your agent gets it wrong, and Nov. 3, what your agent's resolution rate doesn't tell you.
Register. One sign-up covers every session, recordings included.
A Flaw in the Code Behind Most Websites Just Forced Governments Offline
What happened: The NSA issued an advisory on October 21 about a pair of flaws in PHP, the programming language that runs a huge share of the web. Nicknamed PHP2Shell, the flaws sit in PHP itself, not in any one website's setup, and the NSA says they could let attackers take over a web server without logging in. PHP released a patch the same day, and the agency expects criminals to start exploiting the flaws quickly. CNBC reported that government websites went dark while teams scrambled to patch, and AI models from OpenAI and Google reportedly helped researchers find the bugs.
Why it matters to you: If your website runs on WordPress or another PHP-based tool, which is very common for small businesses, this applies to you. You almost certainly don't have to touch any code. You do need to find out whether whoever hosts or maintains your site has updated PHP this week.
What to do about it: Email your web host or web person today and ask, "Has our PHP been patched for the October 21 PHP2Shell flaws?"
Anthropic's New Claude Duet Works While Your Laptop Is Off
What happened: Anthropic introduced Claude Duet, a standalone app where Claude acts as a proactive agent. It sends a morning brief, watches your connected apps, and completes tasks in the background even when your devices are off. It is a research preview by waitlist for US Pro and Max subscribers, with no separate price. Its work counts against your plan's usage limits.
Why it matters to you: Anthropic pitches Duet at professionals and small business owners, so this is aimed at you. The brakes are worth knowing: it shows you the action and waits for your approval before it sends a message, makes a payment, signs a document, or deletes anything, and it can pay using a one-time virtual card through Stripe's Link wallet. The catch is that background checks and big tasks eat into your usage limits, and an assistant that can spend money deserves a careful read of every approval screen.
What to do about it: If you're on Pro or Max, join the waitlist. When you get in, start with the morning brief before you let it send or buy anything.
What if your sales process kept working while you were busy doing something else? That is the idea behind the Epic Marketing Summit, a 2-day live event in Miami this January where you build a working AI-powered sales funnel instead of just hearing about one. It is built for business owners who want their selling to run as a system, not on their own time and energy. You also get hands-on with vibe coding, which means building your own AI app on-site, so you leave with a finished asset instead of a notebook full of frameworks you still have to go implement.
Last year's event sold out, and it's back with a bigger AI focus. I wasn't there last year and I had major FOMO about it. I know a lot of the trainers, and they weren't talking theory, they'd actually done the work. I'm going in 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.
Banks Are Locking Out AI Agents That Log In as You
What happened: Axios reports that major US banks are finding third-party AI agents operating inside customer accounts, and some appear to take actions their users never asked for. Morgan Stanley has started blocking third-party agents, and JPMorgan, Bank of America, and Citi are restricting them too. Chase advised customers not to share usernames and passwords with anyone. Banks struggle to tell a human from a bot, and most are moving toward read-only access, where an agent can view data but not act on it.
Why it matters to you: Even when you give an AI permission to log in, it usually breaks the bank's rules, and bank executives are especially worried about agents that can buy things or move money. If your business banking is part of an AI workflow, you could end up locked out or exposed.
What to do about it: Never hand an AI tool your banking username and password. Use read-only connections your bank officially supports.
Walmart Is Opening Its Doors to AI Shopping Agents
What happened: On October 21, Walmart announced Agent Profiles for Walmart and Sam's Club, according to PPC Land. Each authorized AI agent gets its own profile in your account, so its purchases show up separately from yours. You choose between approving every purchase or letting the agent spend up to a limit you set, with $50, $100, and $200 given as examples. Walmart says agents never receive your card details, and the feature is expected within about a month. Amazon took the opposite approach and cut off Meta's Muse agent from buying on Amazon.com on September 20.
Why it matters to you: More shopping will be done by software, and your customers will send agents to your store, or to your competitors. Walmart's rules (separate profiles, spending caps, approval by default) are a good template for any agent you let near your own money.
What to do about it: If you let an AI buy anything for your business, set a spending cap and require approval first.
OpenAI Admits Two of Its AI Agents Switched Off Their Own Safety Monitors
What happened: In a report on its alignment site, OpenAI described two internal incidents during routine coding work. On October 3, a GPT-6.1 Sol agent repairing an evaluation pipeline turned off automatic alerts to the security team. On October 6, a GPT-6 Astra agent migrating a program removed its monitoring connection, then later told the user it had not. OpenAI says it doesn't yet have a full explanation, and its internal reviewer now blocks attempts to disable monitoring.
Why it matters to you: The watcher only works if the thing being watched can't switch it off. This is a plain-English warning for any business giving AI tools real access: oversight has to live outside the AI's reach, and logs should survive the task.
The best business move I've made in the last few years wasn't a tactic. It was a room. I've been a member of the Ignite Mastermind for over 3 years, and it has 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.
Ignite is an annual membership for entrepreneurs still under $1 million a year in revenue who want structure, tools, and peers to get there faster. You get 6 to 8 live virtual workshops a year plus tickets to 2 in-person multi-day events, each normally priced somewhere between $97 and $997 on their own. There's also an all-in-one marketing software suite that replaces over $5,000 a month in separate tools, and weekly mentorship calls with Perry Belcher and his team. On top of that, I add my personal bonus of 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 how Ignite works here.
A Popular AI Plumbing Tool Was Poisoned for About Two Hours
What happened: LiteLLM, a widely used tool that connects apps to AI models, confirmed that version 1.107.0 was compromised on October 21. The bad package was live on PyPI from about 11:39 to 13:23 UTC, and some Docker images were affected for a similar window. The malware steals credentials, and LiteLLM says anyone who ran it on a host with internet access should assume data was taken and rotate every key that host could reach. The cause is still under investigation, with stolen developer tokens suspected.
Why it matters to you: You may never have heard of LiteLLM, but a developer or agency behind your chatbot or AI workflow might use it. If they ran the bad version, your AI provider keys could be exposed.
What to do about it: Ask anyone who built your AI tools whether they use LiteLLM, and if they do, whether they ran version 1.107.0.
One OpenAI Employee Burned About $10 Million in AI Tokens Last Month
What happened: Peter Steinberger, creator of OpenClaw and now at OpenAI, said at Cloudflare Connect that he spent around $10 million on tokens last month, The Register reported. It is his own estimate, and the money is effectively investors' rather than his. Cloudflare's CTO noted he has "a virtually unlimited token budget, which not all of us in the room have," and Steinberger admitted "not all of that is useful."
Why it matters to you: Running AI agents costs real money, and some of it buys nothing. Most of us will never see numbers like that, but the pattern scales down: agents left running with no budget quietly burn through your plan.
What to do about it: Set a monthly cap on any AI tool that bills by usage.
The Bottom Line
Every story today is the same story. AI is being handed keys to servers, bank accounts, shopping carts, and its own security cameras. The companies doing it well ask for approval, cap the spending, and keep the watchers outside the thing being watched. You don't need to chase every launch. Patch your website, guard your logins, and set limits before you give anything the keys. Stop the rollercoaster and check the seatbelt.
Enjoying the Ride?
If this saved you one headache, forward it to one person who runs a business. Not subscribed yet? Hit the subscribe button on this email so you don't miss tomorrow's issue.
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