Tuesday, October 20, 2026
Today, Washington has decided some artificial intelligence is too dangerous for government computers, lawyers are wondering whether your browser can legally agree to things on your behalf, and investors are beginning to ask whether spending trillions on AI might eventually require earning some money. Meanwhile, the machines are getting more autonomous, which is wonderful news for anyone who has ever wanted their computer to make questionable decisions without the inconvenience of involving them.
Washington Tells DeepSeek to Seek Employment Elsewhere
The Department of Homeland Security has ordered federal civilian agencies to remove products and services associated with DeepSeek, the Chinese artificial intelligence developer.
The concern isn't simply that DeepSeek is Chinese. It's that government agencies may be sending sensitive information through systems whose security, data handling, and potential vulnerabilities raise national security questions.
The order also covers third-party services incorporating DeepSeek technology, which makes things considerably more interesting. An organization might not realize that a product it's using relies on a particular AI model behind the scenes.
This is becoming one of the more important questions in technology: When you use an AI-powered service, do you know where your information goes, who can access it, or what other companies are involved?
For most people, the answer is approximately the same as their understanding of how the refrigerator works.
It gets cold. Please don't complicate things.
Source: Reuters, October 19
Your Browser Has Read the Terms and Conditions. Apparently, You Have Too.
Artificial intelligence is moving beyond answering questions and into operating web browsers on our behalf.
Instead of telling you how to make an appointment, an AI agent might make it. Instead of helping you complete a form, it might complete the form itself.
Which raises an awkward legal question: When the browser clicks “I Agree,” who exactly agreed?
A Reuters legal analysis published today examines how AI-driven browsers challenge existing assumptions about privacy and consent. Laws generally expect people to make informed decisions about sharing information. But autonomous software can click buttons, accept conditions, and disclose data without the user reviewing every step.
Imagine asking your computer to book a haircut and discovering it has also accepted a privacy policy, agreed to promotional messages, and authorized something involving your personal information.
Technically, the computer followed instructions.
Just not necessarily yours.
The technology industry has spent decades trying to eliminate friction from everyday life. Now it may need to reintroduce a little friction so humans can retain some control over what their machines are doing.
Source: Reuters, October 20
Artificial Intelligence Meets Artificial Accounting Optimism
The AI industry has been spending money at a pace that makes government procurement look like careful household budgeting.
Now investors are asking increasingly pointed questions about when all that spending will translate into sustainable profits.
Data centers require enormous amounts of electricity, expensive processors, cooling systems, networking equipment, and financing. The companies building them expect demand for artificial intelligence to keep growing.
But growing demand and making money aren't necessarily the same thing.
Financial markets are showing signs of anxiety about the enormous capital commitments being made across the industry. Investors are scrutinizing debt, financing costs, and whether the revenue generated by AI services can justify the infrastructure required to deliver them.
None of this means artificial intelligence is going away.
It does mean that, sooner or later, someone has to explain the bill.
And preferably not by asking an AI chatbot to produce an optimistic spreadsheet.
Source: Financial Times, October 20
Meta's AI Wants to Help. Some Companies Would Prefer It Didn't.
Meta's AI agent Muse is running into a problem that has less to do with what it can do than with what other companies are willing to let it do.
According to Barron's, organizations including Amazon, JPMorgan Chase, and The New York Times have restricted or blocked Muse over security concerns.
That's an important development in the emerging world of autonomous AI agents.
For these systems to become genuinely useful, they need access to websites, services, and information. But the organizations controlling those resources may not be comfortable allowing automated agents to interact with them freely.
It's rather like hiring a brilliant personal assistant who can do practically anything, then discovering nobody will let the assistant through the front door.
The next great technological competition may not be about which AI is smartest.
It may be about which one everyone else trusts enough to let inside.
Source: Barron's, October 20
The Haywire Takeaway
There's a common thread running through today's stories, and it isn't the usual argument about whether artificial intelligence will replace human beings.
It's about control.
Governments want to control where their information goes. Consumers want to control what their browsers agree to. Investors want to control how much money disappears into data centers. And businesses want to control which automated agents get access to their systems.
For decades, technology companies have promised to do more of our thinking for us.
Now that they're getting better at it, we're discovering how much thinking we'd still prefer to do ourselves.
