Protecting a company with people and AI agents

03How AI changed attacks

AI changed the attacks. This war has no end

What changed the attacks was AI. In one attack, AI carried out 80 to 90 percent of the work on its own. In another case, a single stolen token led to administrator access to the cloud in about three hours.

A cyberwar straight out of the film Summer Wars has begun. There is just one difference from the film: this war has no end. This is the dawn of a new age of war.

05Publishing a fix reveals the weakness

Within hours of the release, exploit code was circulating, and defenders began searching for where PHP was running

Back to the night of October 21. Publishing a fix also tells the world where the weakness is. Research shows that with AI, working exploit code can be built from the diff of a fix in under an hour. This time, too, exploit code began circulating one after another within hours of the release. At that moment, the first thing defenders do is look for where in their own company PHP is running.

06Defense comes down to six functions

Protect and Detect go to specialists. The other four are about how information moves inside the company

Where is it running? Who owns it? Who decides what to fix first, and who needs to be told? Neither firewalls nor detection products can answer these questions.

The NIST Cybersecurity Framework makes this clear. Defense consists of six functions: Govern, Identify, Protect, Detect, Respond and Recover. Protect and Detect are about stopping an intrusion the moment it happens, within seconds to minutes. That is the job of specialist cybersecurity companies and their products. The questions defenders face on the night a PHP fix is published lie with the other four: knowing what you have, who decides, who to tell, and how to recover. That is the flow of information inside the company itself.

07The whole picture is nowhere

PowerPoint, Excel and email are fragments. The whole company is kept nowhere

Many companies still run this flow of information on PowerPoint, Excel and email. Each one is a piece of the whole company, written out for somebody. But the whole itself is kept nowhere. So even on the night the PHP fix was published, defenders start by gathering spreadsheets and emails from all over, just to find out where PHP is running.

Adding dashboards or BI doesn’t remove this searching. They show only part of the numbers. The company as a whole has already grown beyond what a person can take in at once.

09AI agents keep the CITY currentTeleport’s initiative

AI agents redraw the CITY every day, and people make the decisions on top of it

What can keep a CITY up to date every day is not people, but AI agents. Cloud, code, business tools, domains, accounts and permissions. AI agents reread the settings and ledgers and keep redrawing the CITY. A newly opened entrance to the outside, or permissions left behind by someone who has left, show up in the viewer that same day. AI agents keep the whole picture that people could never maintain by hand, and people make the decisions on top of it.

10Replaying the night of October 21Teleport’s initiativeScenario: replaying the night of October 21 at a company that has a CITY

A company with a CITY can find what needs fixing that same night

Let’s replay 11 p.m. on October 21 once more, this time at a company that has a CITY.

When the fix is published, AI agents search the CITY and flag the systems running PHP. Which ones are connected to the outside? Who owns them? Who decides what to fix first? The answers are already in the CITY.

AI agents write up procedures for the owners and priorities for management and deliver them, and people decide the order and make the fixes. As each one is fixed, AI agents verify it again and its flag disappears. Even if an attack has already gotten in, you can decide on the same CITY which data was touched and where to restore from.

No need to gather spreadsheets and emails. That same night, you can get all the way to what needs fixing.

11A team of people and AI agentsTeleport’s initiative

AI agents read, draw and verify; people decide and take responsibility

AI agents read, draw, write up and verify, day and night. People decide, take responsibility and face the customers. Stopping intrusions within seconds to minutes remains, as before, the job of cybersecurity specialists, and CITY gives those specialists the whole picture and the history. The six functions of defense connect on a single CITY.

12Teleport’s CITYTeleport’s initiative

Since August 2026, Teleport has been turning its own company into a CITY

Since August 2026, Teleport has been turning its own company into a CITY. The AI agents’ inventory turned up unused public entrances, permissions still held by people who had left, and notification routes that reached no one. We closed them one by one, recording why.