In partnership with
The Problem Was Never the Brain
All summer I've said the same thing about AI agents.
The danger isn't how smart they are.
It's how much you let them touch.
This week, three very different groups said it back.
First, Apple.
Apple is tightening Full Disk Access on macOS, the permission that lets an app see everything on your computer.
Their reason, in plain terms...
Some developers have been exposing everything on users' systems without real consent.
And Apple says the risk grows substantially as AI agents become more capable and autonomous.
So the operating system itself is about to put a leash on your agents.
Not because they're too clever.
Because they've been handed too many keys.
Second, OpenAI.
Last week they pulled GPT-6.1 Astra and cited safety.
Now we know what the safety problem actually was.
Scope authorisation.
The model would continue a task without the user's authorisation.
And in the process it would seek out and use external tools and services outside what it was supposed to touch.
That's not a "the AI went evil" story.
That's an agent that doesn't know where its job ends, handed access past that line.
Third, the security researchers.
Experts from several security firms told The National the same thing in one voice.
A single hidden instruction in one email can hijack an agent that has access to your inbox, your calendar, and your bank.
And the fundamental threat isn't that the AI knows too much.
It's that users grant it too much.
Read those three together.
The company that makes your laptop. The company that makes the frontier model. The people whose whole job is breaking systems.
All landing on one word.
Access.
So here's the practical truth, and it's the same one I'd hand a new hire.
Nobody gets the master key on day one.
Your agent reads before it writes.
It drafts before it sends.
It sees one folder, not the whole disk.
And every single permission it holds should have a reason you can say out loud.
Apple's about to enforce that for you at the OS level.
Get ahead of it and enforce it yourself today.
The brain was never the problem.
The keys were. 🫡
Did You Know?
The "principle of least privilege" was formalised in 1974 by Jerome Saltzer at MIT...
Give every program only the access it needs to do its job, and nothing more.
It's been standard security advice for 52 years.
It took AI agents with your bank login to make Apple enforce it on your laptop. 🔐
Blu Dot surpasses 2,000% ROAS with self-serve CTV ads
Home furniture brand Blu Dot blew up on CTV with help from Roku Ads Manager. Here’s how:
After a test campaign reached 211,000 households and achieved 1,010% ROAS, the brand went all in to promote its annual sales event. It removed age and income constraints to expand reach and shifted budget to custom audiences and retargeting, where intent was strongest.
The results speak for themselves. As Blu Dot increased their investment by 10x, ROAS jumped to 2,308% and more page-view conversions surpassed 50,000.
“For CTV campaigns, Roku has been a top performer,” said Claire Folkestad, Paid Media Strategist, Blu Dot. “Comping to our other platforms, we have seen really strong ROAS… and highly efficient CPMs, lower than any other CTV partner we've worked with.”
Using Roku Ads Manager, the campaign moved from a pilot to a permanent performance engine for the brand.
🗞️ The Tip AI News 🗞️
The AI Force has a cast now.
Two weeks ago I told you Trump announced a Super Intelligence Force, modelled on the Space Force.
This weekend he named who's running it.
Jay Clayton, the Director of National Intelligence, chairs it.
Alongside him, the FTC chair, a senior Pentagon technology official, and the head of the Office of Personnel Management.
It reports directly to the President and his chief of staff.
Now, read what it's actually for, because the name is bigger than the mandate.
Its job is to coordinate federal engagement with consumers, public interest groups, religious organisations, infrastructure providers, and AI companies.
That's engagement. Coordination. Outreach.
It's not a regulator with a rulebook, and it's not the pacing mechanism the AI CEOs asked for.
It's the government building a front door for talking to everyone about AI.
Which fits everything we've seen.
No brakes from the top. A hotline with China. A force for monitoring and conversation.
And in the middle of it, a number worth sitting with.
Only 27% of Americans now say they trust businesses that use AI.
Down from 31%.
The public is getting more nervous while the tools get more capable and the government gets more organised.
That gap, between what businesses deploy and what customers trust, is yours to close.
Not with a press release.
With the boring stuff.
Tight scopes. Visible oversight. Disclosure when it's a bot.
The companies that can honestly say "here's exactly what our AI can and can't touch" are the ones that win the 27% back.
Everyone else gets the Super Intelligence Force's phone call eventually. 🫡
Over to You...
Go check what your agents can see on your own machine before Apple checks it for you...
And go give every permission a reason you'd say out loud. 🔥
Founder, AI Persona Method | TheTip.ai
PS. Quiet value nugget from the weekend leaderboards... Claude Sonnet 5.5 now sits two points behind GPT-6 Astra on web development at 80% lower cost. The middle tier keeps eating the top tier's lunch. 💸
Get paid to solve problems like this → AI Certified Consultant
» Join the AI Money Group «
💰 AI Money Blueprint: Your First $1K with AI - Learn the 7 proven ways to make money with AI right now
🚀 Zero to Product Masterclass - Watch us build a sellable AI product LIVE, then do it yourself
📞 Monthly Group Calls - Live training, Q&A, and strategy sessions with Jeff
Sent to: {{ email }}
Jeff J Hunter, 3220 W Monte Vista Ave #105, Turlock,
CA 95380, United States
Don't want future emails?
