Three stories this week, and they all point at the same thing. Washington wants to rename AI, Wall Street is lending like the AI buildout can never miss, and Apple is quietly bolting the doors shut on your Mac because AI agents are getting a little too nosy.
01 of 03
The White House Renamed AI and Got a Pinky Promise
On September 29, President Trump signed Executive Order 14434, "Inaugurating the Era of Super Intelligence." It tells every federal agency to stop saying "AI" and start saying "Super Intelligence" or "SI" in correspondence, websites, reports, and policy documents. Old regulations and historical documents are exempt.
Here is the funny part. Section 3 defines Super Intelligence as the exact same thing the law already calls artificial intelligence under 15 U.S.C. 9401(3). Same definition, new sticker. The President's science adviser has 60 days to send over proposed legislative language and say whether the definition should get any bigger.
The next day, September 30, came a White House luncheon and a voluntary pledge called the "Joint Commitment on Frontier Responsibilities." Signers in the room included Elon Musk, Mark Zuckerberg of Meta, Jeff Bezos of Amazon, Sundar Pichai of Google, Jensen Huang of Nvidia, Greg Brockman of OpenAI, Dario Amodei of Anthropic, and Alex Karp of Palantir, plus Trump and House Speaker Mike Johnson.
The pledge lays out four layers of safety checks: internal monitoring during training, in-house testing of safeguards, outside evaluators reviewing the procedures, and independent boards reviewing the reports. Companies pick their own auditors and nobody has to publish results. Trump called it "morally" binding and said "it's almost like a constitution." And per TechCrunch, the signed copy spells it "Unites States" right under his signature.
Here is where I land on it. Renaming the thing does nothing, and I mean that literally, since the order says the definition is identical. The pledge is the part worth watching, and it is weak tea. A safety audit where you hire your own auditor and keep the results in a drawer is a vibe, not oversight. I would take one published eval over a hundred signatures on a misspelled page.
02 of 03
Broadcom Is Borrowing $60 Billion to Build Anthropic's Chips
Bloomberg reported Friday that Broadcom is putting together a $60 billion debt package to fund AI chips and data center gear, with Anthropic as the headline customer. Per Quartz and TradingView, it splits into two pieces. A $42 billion senior secured tranche is being syndicated by banks. An $18 billion junior tranche is led by Blackstone, which is putting in $9 billion of its own money and selling the rest to other investors.
The backdrop is Anthropic's expanded deal with Google and Broadcom for multiple gigawatts of next generation TPU capacity starting in 2027. Reports put Anthropic's commitment at a five year, $125.2 billion infrastructure lease, and Reuters says Anthropic is expected to become Broadcom's largest chip design customer next year. Broadcom set up this financing platform back in June. Meanwhile its stock is down about 1% on the year while the big chip ETF is up around 74%.
What I want to know is who is holding the bag if the music stops. The chipmaker is not just selling chips anymore, it is arranging the loans so the customer can afford them. That is vendor financing, and we have seen that movie before in telecom around 2000. To be fair, Blackstone putting $9 billion of its own cash in the riskiest slice says somebody serious believes the demand is real. I just think everybody should notice how much of this boom now runs on borrowed money.
03 of 03
Apple Is Locking Down Full Disk Access Because of AI Agents
On October 2, Apple posted "Updates to Full Disk Access in macOS" on its developer site. Full Disk Access is the Mac permission that lets an app read basically everything: your files, Mail, Messages, and browsing history. Apple says it will add new controls that require "very explicit user action" before any app gets it, and that the permission should be limited to legit uses like backups.
Apple did not mince words about why. "As AI agents become increasingly capable and autonomous, the risks associated with this level of access will grow substantially," the company wrote. It also said some developers are using the permission in ways that expose users' systems "without users' full knowledge." TechCrunch pointed to Meta's Muse app as one that offers to turn it on. Apple has not said which macOS version gets the change or when.
I am glad somebody said it out loud. Every desktop agent pitch I see wants the keys to the whole house, and most folks click "allow" without reading a word. For communication apps it is worse, because you are also handing over everybody who ever texted you. An agent should get the one folder it needs, not the whole drive. If Apple makes that the default, good. Builders will grumble for a month and then build better.