“Uncomfortable or Exhausting”: What a Human in the Loop Really Means
9-21-2026 (Monday)
Hello, and welcome to The Intentional Brief - your weekly video update on the one big thing in cybersecurity for middle market companies, their investors, and executive teams.
I’m your host, Shay Colson, Managing Partner at Intentional Cybersecurity, and you can find us online at intentionalcyber.com.
Today is Monday, September 21, 2026, and the Strait of Hormuz is still closed.
“Uncomfortable or Exhausting”: What a Human in the Loop Really Means
While we’ve had more news about frontier AI models escaping their sandbox and attacking three innocent companies (as reported exclusively in the Wall Street Journal). This happened back in May, and Google just reported it now - saying “it didn’t consider the hacks to warrant public disclosure—because its model didn’t cause harm to the companies and ended each intrusion immediately upon determining it had hacked a real company rather than a simulated one.”
Or maybe they just needed a marketing boost after their latest cyber model. Who knows?
But that’s not the piece that I want to spend some time on (though the chart in that WSJ piece does show that now all of the large frontier models - from Google, Meta, Anthropic, and OpenAI - have all had a publicly reported version of going rogue).
The piece I think we should spend some time on is another exclusive that dropped that same day from CNN, this outlining how the US moved to intercept “A Chinese ship in the Middle East was transporting components of a nuclear weapons program.” This didn’t, in reality, happen, since the report “according to one of the sources, was “entirely false.” But it also “almost started a war,” the source said. Turns out:
In this particular instance, the analyst queried a chatbot about some intelligence reporting on the ship’s manifest that originated with US Special Operations Command Pacific, based in Hawaii. It was not clear whether the chatbot was a commercially available one or a US government product.
“The internal tools are mostly just copies of the commercial stuff wearing lipstick,” a former senior US official familiar with the AI systems used by military and intelligence analysts.
The bot fused together open-source intelligence with secret signals intelligence in government holdings and reached its fateful conclusion about the material the ship was carrying.
The analyst then used AI again to package the findings into a standard intelligence report — the kind that is trusted by military officials — and disseminated it.
At the same time, also on Friday, the New York Times ran a long piece noting that the Russian government has been targeting Americans with covert online disinformation operations seeking to undermine public confidence in November’s midterm elections.
In each of these cases, we’re seeing things presented as real that simply aren’t - they’re fabricated, whether by the word prediction machines we’re calling AI or by an entity aligned with adversarial governments (who are also, most likely, using AI).
When you hear people talking about keeping a human in the loop, it’s easy to imagine just pressing enter when Claude Code says to (something I’ve been guilty of myself).
The important part, though, is that the human needs to be both able to and empowered and encouraged to think critically with a larger view, and put a stop to things they know aren’t quite right - even if they don’t yet know why.
Some will note how this sort of double-checking will slow down the velocity gains that AI can provide - and that’s true, it is slower than just flipping Claude Code into Auto Mode and letting it rip. But running fast in the wrong direction - especially when that direction is destructive (ships, confidence, code, etc.) - shouldn’t be what any of us are trying to achieve.
Instead, let’s remember that we used to value the views of our colleagues and the people we’ve worked with along the way, building these checks in without even noticing them - they were just inherent in the fabric of the exercise. Amazon even includes this in one of their leadership principles, noting that “Leaders are obligated to respectfully challenge decisions when they disagree, even when doing so is uncomfortable or exhausting.”
And, let me tell you, we’ve all been feeling uncomfortable and exhausted lately, but that’s a sign that we should continue to hold the line, be in the loop, and be willing to disagree and commit. We’re seeing what the alternatives might be, and they’re not good.
Fundraising
From a fundraising perspective, another strong week, with more than $17B in newly committed capital, including:
Goldman Sachs Asset Management raised $9.6b for its ninth flagship PE fund, plus $1.6b for an Asia-focused fund and $500m for related co-investment vehicles; and
Blackstone is targeting $8b for a private credit fund focused on renewables and digital infrastructure - not even counted in this week’s number.
Axios has a report out this morning that Q3 is shaping up to break a streak of six straight quarters with at least $1 trillion in global M&A activity.
Q3 global M&A volume and deal count are down 49% and 39%, respectively, from Q2. The year-over-year totals are down 26% and 41%, respectively.
The U.S. deal dropoff is even more severe, with M&A volume off 61% quarter-over-quarter and deal count down 46.5%. Year-over-year it's fallen by 36.5% and 47%, respectively, including a significant decline in megadeals, which had been inflating the headline numbers.
So far this quarter there only have been nine deals valued at $10 billion or more, compared to 27 such deals last quarter.
Axios notes it could be any number of reasons: High oil prices, stubborn inflation, (correct) expectations of interest rate hikes, lingering SaaSpocalypse fears trickling into other sectors.
“This seems more dramatic than other pauses we've seen," says Matthew Toole, director of deals intelligence for LSEG. "The macro issues are probably causing some buyers to question their models."
Seems like questioning the models is a good theme for the week.
A reminder that you can find links to all the articles we covered below, find back issues of these videos and the written transcripts at intentionalcyber.com.
We’ll see you next week for another edition of the Intentional Brief.
Links
https://hormuzstraitmonitor.com
https://www.cnn.com/2026/09/18/politics/us-military-ai-false-intelligence-china-ship
https://www.nytimes.com/2026/09/18/us/politics/russia-election-disinformation-us-intelligence.html
https://www.axios.com/newsletters/axios-pro-rata-37653e9a-6b87-40f4-a223-c276b1d05265.html