CYBERSECURITY × AI

AI-powered attacks.AI-informed defense.

Many AI-powered attacks use familiar techniques at greater speed and scale.

To counter them, knowledge of both cybersecurity and AI, together with AI-powered real-time detection and defense, is important.

Find your security approach
ATTACKERS + AIANALYZE · AUTOMATEVSDEFENDERS + AIDETECT · DEFEND IN REAL TIME
AI-generated concept art illustrating AI-assisted attack and defense.

CYBERSECURITY × AI

Protect what matters.
Keep business moving.

Your customers' information. Your team's work. The trust you have built.

No one can promise 100% protection from attacks. We combine ways to prevent intrusion, spot trouble early, limit damage and restore operations—around the needs of your business.

Assessment and recommendations tailored to your size, industry and budget

01
Understand

Identify critical work and weaknesses

02
Prepare

Bring people, technology and processes together

03
Continue

Plan for improvement and recovery

URGENT

Latest cyberattack and data leak trends

This is only the beginning.

October attack detections

About 475 million

Cyber Security Cloud recorded its highest total, about 2.5 times the previous October and up around 21 million detections from September.

Kyodo News, 5 November

July–September attack detections

About 1.07 billion

About 11.65 million a day and 135 per second. The company’s detections include requests probing for vulnerabilities.

Cyber Security Cloud, 15 October

Records exposed through October

About 380 million

MM Research Institute includes possible exposures. Around 185 million records were reported exposed in October alone.

MM Research Institute, 4 November
Publicly reported data exposures in JapanAbout 7.6×
Full year 2025About 49.53 million
January–October 2026About 376 million

MM Research Institute compiled reports from Japanese organizations and other public sources that gave a record count.

Automation and distributed attacks

AI-assisted automation may be contributing to rising detections. Researchers also observed attacks distributed across multiple source IP addresses.

Related observations

Risks after a credential leak

Reports show more login IDs and passwords exposed. Stolen credentials can enable unauthorized sign-ins and account takeovers on other services.

THE REALITY

An attack can stop
more than your computers.

Orders, patient care, payments and commitments to business partners. A cyberattack can change an ordinary working day completely.

Ransomware is malicious software that makes data unusable and demands payment. Attackers may also steal information and threaten to publish it.

Reported ransomware incidents in Japan

Full year (January–December)
Reported ransomware incidents in Japan2023:197 / 2024:222 / 2025:226 cases. Reports to police from businesses and other organizations. These are not the total number of incidents or attempted attacks in Japan. Extortion without encryption is counted separately.050100150200250197202322220242262025
View data table
Reported ransomware incidents in Japan · Full year (January–December)
2023197 cases
2024222 cases
2025226 cases

Reports to police from businesses and other organizations. These are not the total number of incidents or attempted attacks in Japan. Extortion without encryption is counted separately. Source 1 ↗

Latest half-year report

First half of 2026 (January–June)

123cases

The highest half-year total on record. It cannot be compared directly with full-year figures. Source 1 ↗

64%

About 64% of reports involved SMEs

79 of 123 reports in the first half of 2026. This is not the percentage of all SMEs that suffered an incident. Source 1 ↗

60%

Investigation and recovery costs of ¥10 million or more

Survey of organizations affected in January–June 2026: 21 of 35 valid responses (60%). These are investigation and recovery costs, not total losses including lost sales. Source 1 ↗

Information checked: November 1, 2026Read the evidence ↗

Published incidents show what we need to prepare for.

We describe published incidents without naming the organizations here. These are not examples of work delivered by Keystone.

Manufacturing & distribution

Incident: September 2025 / Update: February 2026

Orders and shipments stop.

A Japanese food and beverage group lost access to its order and shipping systems. Manual work continued, and the group reported that overall logistics operations returned to normal in February 2026.

Prepare: review connections between sites, administrator access and business recovery.

Source 1 ↗ Source 2 ↗
Healthcare

Published: March 2026

Patient records become unavailable.

A Japanese hospital and care facility experienced a disruption to internal systems. Electronic patient records and other systems could not be viewed, and individual notifications began for people whose information was confirmed to have leaked.

Prepare: check maintenance access, backups and alternative ways to continue care.

Source 1 ↗
Financial services

2025–2026 / Pattern affecting multiple firms

Assets are traded without permission.

Credentials stolen through fake login pages and other methods were used for unauthorized trading in securities accounts. Japan's Financial Services Agency urges stronger identity checks and the use of notifications.

Prepare: strengthen checks on important transactions, detection of unusual activity and alerts.

Source 1 ↗ Source 2 ↗

Photos and visuals are AI-generated illustrations. They do not depict our actual employees, clients or facilities.

SECURITY THAT FITS

Your business is different.
Your security should fit.

Critical work, sensitive information and available staff differ from one organization to another. Choose the starting point closest to yours.

SMALL & MID-SIZED BUSINESS

Security a small team can sustain.

Build arrangements that owners and staff with shared IT responsibilities can manage alongside everyday work. Create evidence that also helps answer business partners' questions.

Shared IT responsibilitiesSupplier security questionnairesAI at work
Start these checks
Photos and visuals are AI-generated illustrations. They do not depict our actual employees, clients or facilities.

YOUR FIRST CHECK

Start with three checks.

0 / 3

Choose an answer for each item to organize the questions you want to discuss. This does not access or assess your systems. Answers are used only on this page and are neither saved nor sent.

01Is someone responsible for disabling accounts when staff or contractors leave?
How to check

Review user lists for email, shared folders and business applications against current employment and contract records.

Possible actions

Disable unused accounts, review permissions, add identity checks and define joiner and leaver procedures.

02Can you actually restore work from your backups?
How to check

Check more than whether a copy exists. Review evidence of a restore to a separate environment and the time needed to resume work.

Possible actions

Use storage that resists deletion, practice restoration and agree which activities and owners take priority.

03Have you agreed what staff may enter into AI tools?
How to check

Review the tools in use, information entered, storage locations, external transfers and contractual settings.

Possible actions

Define AI use rules, approved tools and confidential-data handling, then pilot internal AI search within a limited scope.

This brief checklist does not measure your security level or likelihood of an incident.

What you receive

  • A prioritized improvement list
  • Clear account and operating procedures
  • Recovery and communication exercise records
For technical teams+

We review MFA (identity checks using more than one type of evidence), device updates, SaaS (business software delivered over the internet) settings and restore tests. We also check existing IT providers' contracts and responsibilities to reduce duplication and gaps in ownership.

AI CHANGES BOTH SIDES

AI makes attacks faster.
Keep your defenses evolving.

Convincing fake emails, voice impersonation and faster searches for weaknesses. The AI your own organization uses can also become a new entry point.

Current assessment

AI-related cyber risks entered IPA's 2026 organizational Top 10 at No. 3.

This ranking reflects expert selection, not incident counts or the proportion of victims.

Source 1 ↗
Forecast and assessment through 2027

More efficient, frequent attacks leave less time to respond.

The UK's NCSC assesses that AI will make parts of cyber intrusion more efficient and increase the frequency and intensity of threats. This does not predict a future loss amount or mean that any particular organization will be attacked.

Source 1 ↗
01 / The attacker's approach

AI that deceives people

Natural language or fake voices pressure people to make payments or change access.

Ways to defend

Verify through a separate channel. Require more than one approver for important actions.

02 / The attacker's approach

AI that finds weaknesses

Research into public information and exploitation of known weaknesses become faster.

Ways to defend

Identify internet-facing systems and prepare updates, detection and isolation based on business impact.

03 / The attacker's approach

Attacks that mislead your AI

Malicious instructions hidden in reference documents target secrets or action permissions.

Ways to defend

Separate document access from execution permissions. Check inputs, outputs and actions.

These are practical summaries of threats and defenses, not attack instructions or numerical forecasts.

AI SECURITY BY DESIGN

Let AI help. Keep control of its authority.

We apply our experience using and customizing AI to security design that fits your information and workflows. Start by testing a limited scope: reading, searching and summarizing.

  1. 01

    Check identity and access

    Search only permitted documents

  2. 02

    Link AI assistance to evidence

    Trace answers back to source records

  3. 03

    Approve important actions

    Separate sending, changing and stopping

  4. 04

    Record and review

    Test errors, changes and recovery

This is a proposed architecture. Even with on-premises AI, external connections, storage, administrative access and updates need review. AI output alone is not proof of security.

FROM UNDERSTANDING TO RECOVERY

Beyond installing products.
Build security into everyday work.

We assess your situation and combine the support you need. For specialist testing, monitoring and incident response, proposals define the necessary expertise and service conditions for the engagement.

01

Understand the whole picture

Identify critical work, information, devices and suppliers. Turn findings into a prioritized improvement plan.

Assessment / Asset inventory / Risk evaluation
02

Make intrusion harder

Review accounts, settings, updates and development. Prepare people to recognize attempts to deceive them.

Access & MFA / Configuration / Training
03

Spot problems early

Define what to monitor, who to notify and how to decide what to do.

Monitoring design / Log analysis / AI-assisted investigation
04

Contain and recover

Practice communication, decisions, evidence preservation and restoration.

Incident planning / Restore tests / Exercises

VISIBLE ENGINEERING

Security work you can
see and verify.

Architecture diagrams, configurations, tests and operating records. We focus on deliverables that give business leaders and technical teams evidence for their decisions.

Photos and visuals are AI-generated illustrations. They do not depict our actual employees, clients or facilities.
01

Clear accountability

Agree the roles of owners, approvers, specialists and existing providers, including contact and response hours.

02

Testing within an agreed scope

Define targets, authorization, impact, stop conditions and acceptance criteria before testing.

03

Remaining issues made visible

Separate completed work from unresolved issues, with priorities and clear next owners.

04

Operations and recovery included

Plan reassessment after changes, handover, suspension and restoration.

  1. 01

    Discuss & clarify

    Identify critical work and current concerns.

  2. 02

    Assess & propose

    Agree scope, methods, deliverables, cost and responsibilities.

  3. 03

    Design, implement & test

    Start small and verify configurations and results.

  4. 04

    Handover & improve

    Equip the responsible team and review as needs change.

For technical leaders: six things our proposals clarify+
  • Assets, data flows and company or department boundaries
  • Identities, permissions, AI data sources and action destinations
  • Log timestamps, retention and traceability to original records
  • Test conditions, critical failures and retesting after fixes
  • Ownership and procedures for suspension, recovery and change
  • Scope and service hours, including specialists and partners

Scope and staffing are agreed for each engagement. We do not make blanket guarantees of 24-hour monitoring, emergency response, certification or regulatory compliance.

QUESTIONS & ANSWERS

Frequently asked questions

Technical terms, in everyday language.
I am not sure where to start.+

Start with the work you cannot afford to interrupt, the systems you use and your current IT providers or owners. You do not need to know the technical terms. There is no need to send passwords or real customer data for an initial discussion.

Can we talk if we already have an IT provider or monitoring service?+

Yes. We review current contracts and roles, then consider gaps in design, AI and data integration, testing and operations. Recommendations can build on the systems you already have.

Will AI or a dedicated security product make us secure?+

A product alone is not enough. Security combines use, configuration, permissions, human review, updates and recovery. Even local AI requires checks on external data transfers and administrator access.

How much does it cost, and how long does it take?+

This depends on sites, devices and systems, business impact, assessment methods, specialist needs and the scope of ongoing operations. We define the scope and deliverables, then estimate initial and recurring costs.

What should we do if an incident is happening now?+

Contact your internal incident lead and contracted specialist response service. Do not reset systems or delete records on your own. For medical or production equipment, prioritize decisions by the person responsible for safety. This page is not an emergency reporting service and does not guarantee an immediate response.

Technical terms, in everyday language.+
MFA / Multi-factor authentication
Checking identity with more than a password—for example, a device the person possesses.
Vulnerability
A weakness in software or configuration that an attacker could exploit.
SOC / Security operations center
A team or function that investigates suspicious activity using system records and alerts.
CSIRT / Incident response team
A team or function that coordinates investigation, communication and response when an attack or data leak occurs.
RAG / AI with retrieved information
A system that searches sources such as internal documents and uses them to help AI draft an answer. Accuracy still needs checking.
Prompt injection
An attack that hides malicious instructions in material an AI reads, trying to make it act outside its intended purpose.
Zero trust
An approach that checks users, devices and permissions as needed, rather than assuming something is safe because it is inside the organization.
Digital forensics
Specialist work to preserve evidence from devices and records, then investigate what happened.

LET’S START WITH YOUR BUSINESS

Start with what
you need to protect.

Tell us your size, industry and current concerns. Even if the questions are not yet clear, we can help make the first step concrete.

Contact us on our corporate websiteContact us by emailinfo@keystone-ai.jp

This opens our existing official contact channel. Nothing is sent automatically.

Helpful before a discussion: a list of systems, the scope of existing contracts and work that cannot stop. Please do not send confidential information or passwords.

Sources for the figures, cases and forecasts+

Based on published information available when checked. Figures cover the reports and surveys specified in Japan; they do not represent an individual company's likelihood of an incident. Source links identify the organizations described in the anonymized cases.

Information checked: November 1, 2026

  1. Japan NPA: Cyber threat report 2025 (statistics p.120) ↗2026-03-10
  2. Japan NPA: Cyber threats, Jan–Jun 2026 (main text pp.9,11) ↗2026-09-10
  3. Tokyo Metropolitan Police: Ransomware threats ↗2026-10-16
  4. Japan IPA: Top 10 Information Security Threats 2026 ↗2026-05-21
  5. UK NCSC: Impact of AI on cyber threat to 2027 ↗2025-05-07
  6. UK NCSC: The growing AI cyber divide ↗2025-05-07
  7. Manufacturing case: primary statement, February 2026 ↗2026-02-18
  8. Manufacturing case: updated statement, July 2026 ↗2026-07-17
  9. Healthcare case: primary statement, March 2026 ↗2026-03-26
  10. Japan FSA: Unauthorized access and trading warning ↗2026-10-09
  11. Japan FSA: Reported access and trading incidents, October 2026 ↗2026-10-09