I've been thinking about AI, operational technology (OT), and cybersecurity lately. Conversely I've been having flashbacks to that episode of "The Office" where Ryan Howard tries to push Dunder Mifflin Infinity. Fans will remember this was a customer portal where clients bypass the sales reps and buy paper directly online. Howard eventually threw a chatroom in there for pointless reasons, ignores its various bugs, and doesn't notice customers go back to the sales reps for a more reliable service.

Howard eventually commits corporate fraud by counting regular store sales as online ones to make the website look more successful than it really was. In other words, the meat and potatoes of actual moneymaking was ignored in favor of technology that didn't exactly birth a gold rush.

While some may apply this lesson to the whole of the so-called AI bubble, there is a security hook in that I'm seeing tech companies putting perhaps more effort than is warranted in copying or harping on about Anthropic. I've blathered a lot on this site about Mythos and Project Glasswing, and how OpenAI is very keen to push its own bug hunting cybersecurity AI model extraordinaire.

While I'm sure OpenAI isn't just focusing on Mythos-aping, I did notice it couldn't help cram in some cybersecurity data when it announced its latest GPT‑5.6 series, Sol.

More recently, IBM saw the worst shares drop in its history in advance of underwhelming financial results.

In a letter to investors, CEO Arvind Krishna tried to sweeten the bad news with the following straight after sharing poor preliminary results:

"After the introduction of Mythos, our teams across IBM and Red Hat quickly mobilized to take advantage of an unprecedented opportunity, launching Lightwell. Lightwell is a $5 billion commitment backed by new frontier AI capabilities and a global force of more than 20,000 engineers creating a trusted enterprise clearinghouse to address open source software vulnerabilities."

My first thought was simple: do IBM's investors care about Lightwell? Is the mention of Mythos meant to make them forget the bad news and convince them that IBM will wow the world as much as Anthropic did?

The thing is, Mythos is not ChatGPT. A shareholder can't load it on their phone and experience the "wow" factor.

Also, Project Glasswing is not a moneymaker for Anthropic. It's a beta test of its latest Mythos model, and by extension a good marketing stunt. If anyone's making money from AI security, it's probably OpenAI, if its new joint venture with SoftBank gets off the ground.

My point is that for any company that isn't an AI vendor, Mythos is a Dunder Mifflin Infinity in sheep's clothing.

If you want to know where the real gold rush in cybersecurity is right now, you have to look at something a bit more everyday, perhaps a bit drab in nature. Think of Dwight and Stanley in "The Office" instead of Ryan letting his hair down at VIP parties in New York City.

You have to look at something that isn't OTT (over the top) at all.

I'm talking about operational technology security.

Over the T

Earlier this year I covered ServiceNow splashing $7.7 billion on security firm Armis, some short months after Mitsubishi Electric bought Armis rival Nozomi Networks for a rumored $880 million.

June then saw consulting giant Accenture gobble up Dragos for $4.1 billion.

The connection between all these firms? They're all OT security vendors. Some smaller European deals set the theme: OT security is defining cybersecurity M&As in 2026, And nobody's really noticed because A) Mythos; B) Accenture got hacked, distracting everyone from its Dragos deal; and C) OT is as fashionable as old Creed Bratton lurking in the corner of the office.

But with so many industrial environments running decades-old technology never designed with security in mind, the sector is overdue a digital revamp. It may not be flashy, but we're talking about a threat to some of the most critical sectors in society: energy, water, transport, and manufacturing.

And while new Accenture darling Dragos may be on Project Glasswing, Anthropic's security focus remains on operating systems, browsers, and open-source projects.

Also, while there have been various deals around identity solutions to protect enterprises from shadowy AI agents, they've all been scooped up by fellow cybersececurity vendors. It's all within the security ecosystem, but, as you can see, OT security is seeing interest from enterprise players, consultancies, and manufacturing giants.

There just aren't enough AI agents yet out there on the network for people to start worrying. Certainly not as many as there are industrial businesses, which are one step away from cyber-disaster.

In fact, if "The Office" was made today, you just know Ryan Howard would be flogging a paper-centric AI agent.

And it'd be called Dunder Mythos GPT.