In partnership with
Hey friends 👋
Big one today: security researchers found what they're calling the first fully autonomous AI ransomware attack, except the details tell a messier story than the headline suggests. Also, California just signed a massive Claude deal with the federal government simultaneously flagging Anthropic as a supply-chain risk. Let's dig in.
Researchers Found the First "Fully Autonomous" AI Ransomware Attack. The Details Tell a Messier Story.
Sysdig's threat research team spent July 4 through July 6 digging into something called JADEPUFFER, a ransomware operation it's calling the first documented end-to-end autonomous AI attack. Six hundred payloads deployed. No human at the keyboard, supposedly.
But "fully autonomous" is one of those phrases that sounds airtight until you start asking questions. Did the AI pick the target? Did it decide when to strike? Or did a person set the parameters, kick things off, and let the AI handle execution while still calling the actual shots? Those are wildly different threat models wearing the same headline.
This matters because "autonomous AI attack" is going to be the scariest phrase in cybersecurity for the next few years, and it's going to get used loosely. If the bar for "fully autonomous" turns out to be "AI executed the technical steps while a human still made the calls," that changes how worried you should be, and it changes what defenses actually work. Worth watching which one this actually was once more details land.
Try the AI that knows your customers. No commitment.
Most platform evaluations start with a demo request and end three weeks later in a conference room. This one takes 15 minutes and puts you directly inside Gladly's interface — navigating it on your own terms.
See how AI surfaces real-time customer context before a conversation starts. Watch how a single conversation thread pulls in purchase history, channel history, and account details without a handoff.
No installation. No commitment. Start the interactive demo and see the platform for yourself.
California Just Gave Every State Agency a Claude Discount. The Feds Just Called Anthropic a Security Risk.
Gavin Newsom signed a deal this week giving every California state agency, city, and county access to Claude at 50% off. Same week, the federal government designated Anthropic a "supply-chain risk."
Sit with that for a second. A state government just went all-in on the exact company Washington is flagging as risky enough to warrant that label. That's not a subtle disagreement, that's two arms of American government looking at the same company and landing on opposite conclusions about whether it's safe to build public infrastructure on top of.
Part of this is just politics, California and the current federal administration don't exactly see eye to eye these days, and an AI deal was always going to become a proxy fight eventually. But part of it is a genuinely unresolved question: is Anthropic a trusted infrastructure partner or a risk that needs monitoring? Usually you get one answer. Right now you're getting both, from the government, at the same time.
A few more things worth knowing
Gemini 3.5 Pro still hasn't shipped. Sundar Pichai promised it at I/O in May, Google now says July, and the stated reason is enterprise testers flagging excessive token burn on long agentic tasks. Google shipped two image models instead to fill the gap.
SK Hynix is reportedly eyeing a $28 billion US IPO. It makes the HBM memory chips that go into every Nvidia Blackwell GPU, so this would be one of the more direct ways to invest in AI infrastructure specifically rather than the model layer.
Anthropic signed a $19 billion data center lease with TeraWulf, a nuclear- and hydro-powered compute provider. Adds to a growing pile of leases as Anthropic reportedly preps an October IPO filing.
Anthropic has reportedly overtaken OpenAI on revenue, per financial reporting this week. Worth a full item on its own once the numbers firm up.
That's what stood out to me today. Reply and tell me what caught your eye, I read everything.
Talk tomorrow,
Hatman 🎩