New AI platforms hand hackers powerful new tools for cracking cybersecurity

Bundles of currency on display at the Federal Reserve Bank's Money Museum, 230 S. LaSalle St. 

     Everyone has a morning routine. With me, I open my eyes, muse darkly upon my life, then flee upstairs to my office — even before coffee is made or the dog walked — blast out the day's blog post to my hearty band of followers, then ritualistically log into a financial service company to check on my 401(k). A moment emotionally somewhere between Scrooge McDuck going down into his vault to roll around on his piles of gold and a castaway in a rubber raft checking the amount of water left in the jug. It takes a minute.
     Except one Sunday morning a few weeks ago. I plug in my username and password. Nothing. A second, more careful try. It warns me, in red letters, that a few more such attempts and I will be locked out.
     Not wanting that, I hit "Forgot password." It asks for my birthday, my email and the last four digits of my Social Security number. Normally, you never share that information. But this wasn't something over the transom; I'd logged into my 401(k) site. I plug in the info.
     No account associated with me. I try again. Nothing.
     A few more attempts, with growing alarm, that I'll spare you. In brief, my 401(k) account, with my entire nest egg needed for looming retirement, built over decades, the provisions that must sustain us on our one way journey into the dark woods of decline, had simply vanished.
     Money, as you know, is no longer bullion slumbering in vaults or even stacks of fresh currency, but mostly bundles of electrons flitting through systems of unfathomable complexity. You buy a pack of gum, tap your phone to a contactless payment terminal, and great institutions briefly kiss. Visa slips Walmart $3 and debits your account. You get a pack of Hubba Bubba Sour Blue Raspberry.
     We hardly even think about it. But maybe we should. While we're used to the idea of endless legions of scammers assailing us through every mode of communications short of semaphore flag, the latest and most ominous twist is coming from a new weapon of immense power that is already derailing modern life: artificial intelligence.
     Yes, AI. The thing that keeps trying to summarize your email. That your kid uses to write his report on Cotton Mather instead of actually doing the work and learning something. AI is so incredibly powerful, not only does it produce videos of obese porch pirates getting their faces painted with blue dye, but it can code/write computer programs.
     Or crack them. A story that might have gotten lost in the whirl of general disaster is that on April 7, Anthropic, an AI company that started five years ago and is now worth $380 billion, provided a preview of its Mythos AI model to 40 Big tech giants — Apple, Google, Microsoft, JPMorganChase. The reason for this effort, dubbed Project Glasswing, is because AI can cut through cybersecurity like a hot knife through butter, thwarting encryption, discovering hidden vulnerabilities that escaped notice for years.

     So Anthropic is giving the biggest players a chance to fix their heretofore undetected flaws before Mythos is available to the general public, one of whom might decide to type, "Drain Neil Steinberg's 401(k) and transfer the contents to my Apple Wallet."

To continue reading, click here.