[Roundtable Vol. 5] "Hackers Don't Break In, They Log In." Cyber Threats in the AI Era, and a Practical Answer to the Growing Need to Eliminate Orphaned Accounts and Unused Access Rights

Updated Tactna Editorial8 min read

Note: This article was originally written in Japanese and translated into English. Some nuances may differ from the original text.

TL;DR

  • A paradigm shift in threats: Modern hackers don't "break in" to systems. They boldly "log in" by abusing neglected, legitimate IDs. That is exactly why leaving unused IDs unattended is dangerous.
  • New risks in the AI era: Automated attacks and runaway AI agents with no ethical judgment have made leaving unnecessary access rights unattended a more critical risk than ever.
  • The limits of centralized management, and where the world is heading: It is physically impossible for a headquarters IT department to keep track of external partners' resignations and transfers. Meanwhile, the world is moving toward strict Identity and Access Management that eliminates orphaned accounts.
  • Tactna's practical answer: Tactna's "autonomous decentralized" approach lets external partners' own administrators manage their members, reducing operational burden while preventing unused IDs and access rights from being left unattended.

Introduction

Hello, this is Aso from the Corporate team.

We have some news: our company has been selected for the Silicon Valley course backed by the global venture capital firm 500 Global, under the "J-StarX" program promoted by Japan's Ministry of Economy, Trade and Industry (METI). (We hadn't officially announced it yet!) Our CEO, Sudo, is in Silicon Valley right now, and we held this roundtable in the middle of his business trip.

In Japan, corporate data breaches have been making headlines one after another. But what Sudo is experiencing firsthand, where top talent from around the world gathers, is the rapidly evolving threat of AI, and the solid potential of Tactna as a defense platform to counter it.

In our previous post, we talked about how neglected "zombie IDs" (accounts that are no longer used but remain active) are straining operations on the front lines. This time, we go into an even more serious side of the problem: security threats and compliance requirements.

Under the striking theme "Hackers don't break in, they log in," Sudo in Silicon Valley joined Sanada, Ezaki, and Iwakura, who work on Tactna's development and delivery, for a deep dive.

1. "Hackers Don't Break In, They Log In": The Latest Cyber Attack Trends

Sudo: An article from Cybersecurity Insiders has been a hot topic in Silicon Valley lately, and it makes the same point: today's hackers don't exploit system vulnerabilities to "break in." They boldly "log in" with legitimate IDs and tokens. According to IBM's 2025 Cost of a Data Breach Report, the average cost of a data breach in the U.S. has reached a record $10.22 million. That's a frightening number. We're now in an era where we have to prepare not only for attacks that exploit vulnerabilities, but also for attacks that abuse legitimate credentials.

Ezaki: As the infrastructure person here, this change in the intrusion route is a real headache. It's no longer enough to put a strong firewall in place and call it a day. Once an attacker gets in through a legitimate ID, the most trusted route there is, ID management itself becomes the lifeline of the infrastructure from a defender's point of view.

2. The Risks of Excessive Data Retention, and the Wave of Strict Management of Unused IDs and Access Rights

Sanada: From a development perspective, one big problem is that companies collect far more data than they need, for marketing and other purposes. I want to stress how important it is not to hold on to original data once it's no longer needed, after confirming any legal and contractual retention requirements.

Iwakura: From an operations standpoint, I completely agree with Sanada. Leaving unused IDs and data in a system is like leaving assets (or rather, liabilities) lying around and telling attackers, "Please, target this." It's a major factor in amplifying the damage when an incident occurs.

Aso: That's true. Even in recent news, a survey by Teikoku Databank, a major Japanese credit research firm, found that about 30% of large corporations have experienced a data breach or a possible one. Across all responding companies, 14.8% answered "I don't know," and apparently some said it was because they "don't have experts."

Ezaki: Given that situation, the importance of promptly disabling unused IDs and access rights (so-called zombie IDs) instead of leaving them unattended is growing across society.

3. The Risk of Reckless AI Without Ethics, and the Limits of IT Departments

Sudo: What makes it even more troublesome is that AI is being abused to automate and speed up attacks like password-list attacks. Leaving unused IDs unattended has become a more critical risk than ever. On top of that, instructions alone can't guarantee that AI agents will operate safely, so there's also a risk that they could unintentionally break into other companies' systems, leak data, or destroy it.

Sanada: To add a technical point: AI agents sometimes don't grasp unspoken rules like "reading the room" or "surely this shouldn't be deleted." They simply try to complete the task as efficiently as possible within the permissions they've been given.

Sudo: Exactly. Incidents like an AI agent breaking into Hugging Face are actually happening. So are cases like PocketOS, a SaaS company running on Railway, where an AI told to fix an authentication error in a staging environment accidentally deleted the production database. Railway kept separate disaster-recovery backups, so the data was restored. But if unnecessary permissions (that is, neglected access rights) or more permissions than needed are left in place, AI can make full use of them and cause an unexpected major accident.

Ezaki: What scares me from an infrastructure perspective is the speed and breadth of what AI can do. Even mistakes that a human would notice and stop, AI can turn into destroyed systems or extracted data across a wide area in an instant. That's exactly why strict control over "who has which permissions" is demanded even more severely in the AI era.

Iwakura: That said, AI can also be a powerful tool for streamlining administrative work. Like the agent mode in the Auth0 CLI, handing administrators' manual tasks over to AI agents has the potential to dramatically cut labor costs.

Ezaki: Still, I think there is still a bit of a hurdle before AI agents spread across entire companies. If you try to adopt AI agents seriously for business, even SaaS-type tools can cost tens of thousands to hundreds of thousands of yen per month (roughly US$100 to a few thousand), and full-scale operation often costs even more. So there's still a barrier to widespread adoption by non-IT employees.

Aso: But if it runs in the background without users even noticing, like the AI features on smartphones, wouldn't it spread quickly?

Sudo: No doubt about it. You can see that just from the quality of AI-generated creative work. And as electricity demand grows toward 2030 and next-generation energy lowers infrastructure costs, mainstream adoption is sure to expand.

Aso: In an era where AI moves autonomously like that, it seems difficult for a headquarters IT department to keep track, in real time, of external partners' resignations and transfers and delete their IDs on its own.

Iwakura: Exactly. What we often hear in the field during delivery is that the ID issuance process for external partners is still analog: "Email an Excel application form, IT registers the account by hand, and sends the password in a separate email." Onboarding might work with that, but reports of resignations and transfers don't reach IT in a timely manner. As a result, accounts left unattended for months, or even years, pile up. Something is bound to slip through somewhere, and those accounts become prime targets for hackers.

4. How Tactna Differs from Existing IGA, and Its "Autonomous Decentralized" Defense Platform

Aso: That's where Tactna's approach of delegating management to the external party hits the mark, right? But can't existing ID management tools do that?

Sudo: Existing enterprise IGA (Identity Governance and Administration) products like SailPoint and Saviynt are excellent tools that also offer external user management and delegated administration. Tactna, on the other hand, focuses on providing an integrated service for external partners that covers authentication, organization management, usage rights management, and an administrator portal.

Iwakura: So instead of provisioning directly to each application, we manage the lifecycle of organizations, members, and usage rights on Tactna, and we delegate the actual operations (who to add, who to remove) to the external partner's administrator, who knows their team's situation best. Because we leave those decisions to them through a self-service portal, our biggest strength is that Tactna can be a foundation that reduces the number of unused IDs and access rights left unattended.

Ezaki: Looking at it from an architectural standpoint, Tactna's autonomous decentralized model makes total sense. Instead of managing everything centrally, it hands management authority to the field while keeping the management status and operation history visible centrally. That lightens the burden on the people doing the managing while keeping security intact. Security shouldn't be built on a belief in human goodness. You have to build a Zero Trust architecture on the premise that data will leak and systems will be breached, and Tactna can serve as that foundation.

Sudo: That value has already been proven. In one customer's case, about 30% of their external IDs turned out to be zombie IDs. After they introduced Tactna and ran automated inventories (including some custom implementation), they told us that, in their experience, the time spent on ID management and inventory dropped by about 70%. (Read the case study here.)

Aso: It's reassuring to have a solid track record. By the way, Sudo, you're in Silicon Valley right now. Do you feel Tactna's approach holds up on the front lines of AI?

Sudo: Yes. Being here has actually made me even more convinced of its usefulness. I can feel firsthand that the "Made in Japan security" brand, meticulous and robust, is a big strength when it comes to giving people peace of mind.

Sanada: That's exciting! Precisely because the whole world is worried about runaway AI and increasingly advanced hacking, a Japan-born architecture is a great fit.

Sudo: Exactly. In the era ahead, when strict Identity and Access Management is demanded and AI roams through our systems, let's promote Tactna to markets in Japan and abroad, not just as an operational efficiency tool, but as a robust defense platform supporting Japanese security.

Editor's Note

What did you think? In an era when hackers have shifted from "breaking in" to "logging in," and AI roams through systems, leaving unused IDs unattended is a critical risk for companies. As society's demand for strict Identity and Access Management grows, this roundtable once again highlighted the importance of autonomous decentralized ID management, which keeps security intact without exhausting IT departments. If you're struggling with external partner ID management, please consider Tactna's solutions.


Keep partner IDs
up to date.

  • Audit log on every plan
  • 70% less admin work in a customer case