This was one of the busiest weeks in AI in a long time. OpenAI put GPT-6 in the hands of every ChatGPT user, the cost of running AI dropped across the board, and a wave of new tools landed that touch everything from your paperwork to your customers’ Instagram DMs. Here’s what actually matters if you run a business — and the five questions worth asking before your competitors do.
The 5 questions at a glance
- Has my business just become a target for AI-powered cyberattacks — and what can I actually do about it?
- Can medical spas, dental offices, and law firms finally use AI without client data leaving the building?
- What does Meta’s new booking agent inside WhatsApp and Instagram mean for restaurants, salons, and home-services companies?
- For accountants, insurance agents, and real estate offices — is the paperwork finally something AI can handle?
- Did AI just get cheaper — and for contractors and manufacturers, is a specialized model now better than the big names?
1. Has my business just become a target for AI-powered cyberattacks — and what can I actually do about it?
This week OpenAI did something unusual. It launched GPT-6 (codenamed “Astra”) to every ChatGPT user and openly admitted the model scored at its highest internal cyber-capability tier — meaning that in testing, it could find and chain together software vulnerabilities on its own. Google shipped a model called Gemini 3.8 Flash Cyber the same week, and Anthropic released a locked-down “Mythos” model that only vetted defenders can access. More than 100 companies signed a public warning that self-directed AI attacks could soon outpace human defenders.
Translation for a business owner: the tools that can probe your website, your booking system, and your customer database for weaknesses just got dramatically more capable — and cheaper for bad actors to use. If you run a retail or e-commerce shop, a medical or dental practice, or any professional-services firm sitting on client data, you’re squarely in range. This isn’t a reason to panic; it’s a reason to finally do the basics you’ve been putting off.
The good news is the same wave gives you affordable defense. Google’s new Flash Cyber model finds and patches vulnerabilities at a fraction of what a security consultant costs, which means “have someone check our systems” is finally a line item you can afford. Ask your web developer or IT person when they last audited your site and payment flow, turn on multi-factor authentication everywhere, and make sure one specific person actually owns security at your company.
2. Can medical spas, dental offices, and law firms finally use AI without client data leaving the building?
For years, the honest answer to “can we use AI in our practice?” for anyone handling sensitive records was “not really” — because using most AI tools meant sending client files to someone else’s cloud. This week that changed. Perplexity launched “Hybrid Compute” on Mac, which keeps sensitive files on your own computer and only sends the non-sensitive parts to the cloud, paired with a built-in detector that flags personal information before it ever leaves the building.
If you run a medical spa, a dental or medical office, a law firm, or an accounting practice, this is the development to watch. Anthropic also rolled out “zero data retention” options for businesses this week, and Cohere is building an entire “sovereign AI” business around keeping data inside your own country. The theme is the same everywhere: where your data physically lives is now a feature you can choose, not a compromise you’re forced to accept.
Practically, if a compliance worry — HIPAA, client confidentiality — has stalled an AI project, it’s worth reopening. Start with a low-stakes workflow, like drafting appointment reminders or summarizing non-sensitive intake notes, on a tool that keeps data local, and prove it out before you touch anything protected. The blocker that stopped you six months ago may simply be gone.
3. What does Meta’s new booking agent inside WhatsApp and Instagram mean for restaurants, salons, and home-services companies?
Meta began rolling out “Hatch” this week — an AI agent that lives inside WhatsApp and Instagram and can actually do things for people: make a purchase, book a restaurant, complete a task, all inside a chat. Meta reaches billions of people in those apps, so this puts a transactional AI in more pockets than any business software ever could.
If you own a restaurant, a salon or med spa, a home-services company, or a real estate office — anywhere customers book appointments or place orders — pay close attention to how people find and book you. When a customer can type “book me a table Friday” or “find me a plumber” inside Instagram and an agent handles it, the businesses that show up cleanly — accurate hours, current menus or services, instant-bookable listings, fast replies — will win those handoffs. The ones with a stale Google listing and a “we’ll call you back” form will quietly lose them.
You don’t need to build anything fancy. Make sure your business information is accurate and consistent everywhere an agent might read it, that your booking link works on a phone in two taps, and that someone responds to DMs quickly. The shift is that a customer’s first interaction may now be with an AI, not with you — so your online presence has to make that AI’s job easy.
4. For accountants, insurance agents, and real estate offices — is the paperwork finally something AI can handle?
Mistral released “OCR 4” this week, and it’s a real jump for anyone who drowns in paperwork. Independent reviewers preferred it over the leading document-reading systems, and its most useful trick is that it tells you how confident it is about each thing it reads. Instead of quietly guessing on a smudged number, it flags the uncertain bits so a human can double-check them.
That “knows when it’s unsure” feature is exactly what has kept accountants, bookkeepers, insurance agents, real estate offices, and property managers from trusting automation with invoices, claims, contracts, and applications. If the software might silently misread a dollar amount, you can’t use it. A tool that raises its hand and says “check this one” can be wired into your workflow safely — the machine handles the 90% it’s sure about and routes the rest to a person.
If your team spends hours keying in data from PDFs, scanned forms, or contracts, this is worth a pilot. Pick one painful, document-heavy process — matching invoices to purchase orders, or pulling data off insurance forms — and test a modern document AI on last month’s real paperwork, then measure the hours saved. This is often the single highest-return place a small business can put AI to work, and it just got more reliable.
5. Did AI just get cheaper — and for contractors and manufacturers, is a specialized model now better than the big names?
Two things happened this week that should make you re-check what you’re paying. First, DeepSeek released a free, “open-weight” model (V4-Pro) that performs near the level of the expensive market leaders, and Anthropic cut one of its key pricing components by 75%. When free-to-run models get “good enough,” the paid tools have to compete on price — and that pressure eventually reaches the subscriptions and software you actually buy.
Second, specialization is starting to beat raw size. xAI is about to release Grok 4.7, trained partly on SpaceX’s engineering data specifically to be better at technical and manufacturing tasks. For a general question, the big-name models are largely interchangeable. But if you run an HVAC company, a contracting or manufacturing business, or a machine shop, a model tuned for technical work may soon outperform the famous ones on the jobs you actually care about — sizing, specs, troubleshooting, and engineering math.
The takeaway isn’t to switch tools every week. It’s to stop assuming the most famous or most expensive AI is automatically the right one for you. Once a quarter, look at what you’re paying for AI, what you’re using it for, and whether a cheaper or more specialized option now does the job. Prices and capabilities are moving fast enough that “we set this up last year” is no longer a good reason to keep overpaying.
What this all means
Underneath a noisy week are three shifts worth remembering. AI security now cuts both ways — the same power that makes attacks cheaper also makes defense affordable, so the basics matter more than ever. Where your data lives has become something you can control, which reopens AI for the practices that had to sit it out. And the cost and shape of AI are changing fast enough that the smart move is to reassess regularly rather than set it and forget it. None of this requires a big budget — it requires knowing which questions to ask.
Not sure how any of this applies to your specific business? That’s exactly what our free audit is designed to answer. In 30 minutes, we’ll map out where AI can realistically save you time, generate leads, or give you visibility you don’t currently have — no jargon, no pressure. Book your free audit →