Stage 3 of 3 · Act

Which defenses stop AI-driven attacks, and which to fund first

Attackers now use AI agents to find a way in and spread within minutes. Here are the defenses you can put in place, from basic controls to AI tools of your own, and a simple way to decide which fit your company.

  • 9 min read
  • For CEOs, boards and CIOs
  • Technical parts marked

Board questions this part answers

  • What can each of our AI agents actually do, and can we stop it?
  • Which risks have we accepted, and who owns them?

The 6 jobs of cyber security, and the frameworks for each

The most widely used map is NIST's Cybersecurity Framework 2.0. It sorts security into 6 jobs, and the other frameworks your team or auditor mentions fit inside them. Tap a job to see what it asks of leaders and which frameworks go deeper on it.

Tap a job on the ring, then switch on the AI layer

Govern sits at the center and steers the other 5.

Govern

Decide who owns cyber risk and how much of it you accept.

What it asks of leaders

Name an executive owner, set a risk appetite the board signs, and review it every quarter. This is the job added in 2024, and it puts cyber risk on the board's agenda.

Frameworks that go deeper

  • NIST CSF 2.0
  • ISO/IEC 27001
  • ISO/IEC 42001 for AI

In this report Board decision (90-day plan)

Go deeperWhich framework is for what
Which framework is for what
FrameworkThe question it answersWho uses itWhen you need it
NIST CSF 2.0Are we covering all six jobs?Boards and security leadersAs the shared map for everyone
CIS Controls v8.1What exactly should we do first?IT teamsTo turn the map into a task list, in 3 levels
ISO/IEC 27001Can we prove it to others?Companies selling to large clients or abroadWhen customers or regulators ask for certification
Zero TrustWhat if someone is already inside?IT architectsWhen moving to cloud or opening systems to partners and agents
OWASP Top 10 for agentic appsHow do AI agents get abused?Teams building or buying agentsBefore any agent can act on its own
MITRE ATT&CK and ATLASHow do attackers actually work?Security teams and providersTo test detection against real attacker behavior
ISO/IEC 42001Do we manage AI responsibly?Companies with many AI usesWhen AI becomes part of core operations
UU PDP, OJK and BI rulesWhat does the law require of us?Legal, compliance and ITAlways, in Indonesia

CIS Controls map onto NIST CSF 2.0, and OWASP and ATLAS map onto both, so your team can start with one and add the others later. ,,,

5 levels of defense, and the order to fund them

Each level builds on the one below. Most companies should finish level 1 before buying anything at level 3. Tap a level to see what it stops and what it takes.

Tap a step to see what each level stops

Level 1

Essentials

Close the doors most attacks use

What it stops

Stolen passwords, fake sign-in pages, known flaws, payment fraud by email, and ransomware that would otherwise wipe out your only copy.

  • Multi-factor sign-in on email, admin and remote access. In a large study of accounts showing suspicious activity, it cut the risk of takeover by 99.22%.
  • Critical flaws fixed within days, and a list of every system that faces the internet.
  • Offline backups, tested by restoring them.
  • A callback and a second approver before any change to bank details or any large payment.
PeopleBudget

What it takes

One capable IT lead and a few weeks. Mostly settings and habits, with little new to buy.

Before you start

Nothing. Every company starts here.

Owned by Head of IT, with the CFO for payments

UK companies certified to a similar set of basic controls make 92% fewer cyber insurance claims.

8 kinds of AI that defend you

Each card is a type of tool, and most security vendors now sell several types. Tap one to see what it does, the evidence so far, and the question to ask before you buy.

Tap a kind of AI to open it

AI email and payment-fraud screening

Businesses reported US$3.05bn lost to email fraud to the FBI in 2025.

What it does

Learns how your suppliers and executives normally write and what they normally ask for. A request to change bank details from a lookalike address, or in an unusual tone, is held for checking.

Where it falls short, and what to ask

It cannot stop a fraud that arrives by phone or WhatsApp. Keep the callback rule. Ask how many of last month's held emails were real fraud, and how many were false alarms.

Useful from Level 3, for any company that pays suppliers by transfer

Which attack responses AI may take alone, and which need a person

Let AI act alone only where the action is quick to undo. Agree these lines before you switch anything on, and review them every quarter.

  • 29 min

    the average time criminals took to spread to other machines in 2025.

  • 14 days

    the median time an attacker stayed unnoticed in 2025.

Who may do what during an attack
ActionAI aloneAI proposes, a person approvesPeople only
Sort and summarize alertsYes
Block a known malicious email or websiteYes
Isolate one laptopYes, if it can be undone in minutes
Suspend an employee's accountYes
Shut down a server or production lineYes
Cut off a supplier's connectionYes, out of hoursYes, in hours
Tell customers, regulators or the policeYes

The same thinking applies to your business AI agents. The authority ladder in the AI agents chapter sets how much each one may do.

6 questions to find your level, and who should run it

Answer 6 questions about your company. You get the level to aim for this year, a way to resource it, and 3 first moves. Nothing you enter leaves this page.

Answer 6 quick questions about your company

Question 1 of 6How many people work in the company?
Go deeperHow the result is worked out

Company size and IT team each score 0 to 3. Level 2 applies when either scores 1 or more, when you are regulated, or when the business stops within hours. Level 3 applies when the two together reach 3, or when a regulated company has 50 people or more. Level 4 applies when you have dedicated security staff and either 250 people or more or a regulator. Level 5 is added whenever agents take actions. The way to resource it follows your IT team. The moves start at level 1 unless all three basics are in place, then follow the levels upward. When your agents take actions, securing them becomes the second move. Treat it as a starting point for a conversation with your CIO. It is not professional advice.

7 questions to ask before you buy any security tool

Every vendor in these categories will say its product uses AI. These questions show what it actually does, and what you get if you leave.

Board questions this part answers

  • What can each of our AI agents actually do, and can we stop it?
  • Which risks have we accepted, and who owns them?

Questions leaders ask

What is the best cyber security framework for a company?

Use NIST's Cybersecurity Framework 2.0 as the map. Other frameworks map onto its 6 jobs: govern, identify, protect, detect, respond and recover. Use CIS Controls to turn it into tasks, and ISO/IEC 27001 when customers ask for certification. ,

Can AI protect a company from cyber attacks?

AI helps defenders sort alerts, screen email and contain attacks faster, and companies using security AI and automation saved US$1.93m per breach on average in IBM's 2026 study. It does not replace the basics: multi-factor sign-in, patching and tested backups still close the most common ways in. ,,

What is MDR in cyber security?

Managed detection and response is a service where a provider's team watches your systems 24/7, investigates alerts and can contain an attack, such as isolating a laptop, on your behalf. It suits companies without their own round-the-clock security team.

Should AI be allowed to respond to cyber attacks on its own?

Only for actions that are low-risk and quick to undo, such as blocking a known bad email or isolating one laptop. Suspending accounts, shutting down systems and informing regulators should need a person. Agree these lines in writing before switching automation on.

How do we secure our own AI agents?

Keep a register of every agent with an owner, give each its own identity and the least access its job needs, check its inputs and actions through a guardrail or gateway, and require a person to approve payments, deletions and data leaving the company. The OWASP Top 10 for agentic applications lists the risks to test against. ,

Want a second pair of eyes on your 90 days?

Red Asia Group's technology companies run security operations and testing for enterprises in Indonesia. If you want help turning this report into your own plan, start with a conversation.

Sources for this part22 sources
  1. [3]

    IBM · Study: one in four malicious breaches are AI-enabled, costing companies $6 million on average

    29 July 2026. 602 organizations breached between March 2025 and February 2026.

    Open the original
  2. [4]

    IBM and Ponemon Institute · Cost of a Data Breach Report 2026

    July 2026. Global average US$4.99m. The ASEAN figure (US$4.12m) comes from a sample of 26 organizations.

    Open the original
  3. [5]

    FBI Internet Crime Complaint Center · 2025 Internet Crime Report

    April 2026. Complaints made to the FBI, including consumer fraud. Business email compromise: US$3.05bn from 24,768 complaints.

    Open the original
  4. [7]

    CrowdStrike · 2026 Global Threat Report

    24 February 2026. Breakout time is how long criminals take to move from the first machine to others, observed in CrowdStrike's own cases.

    Open the original
  5. [14]

    CISA, FBI and partners · Scattered Spider joint advisory (AA23-320A)

    Revised 29 July 2025. Describes criminals calling IT help desks to have passwords reset and sign-in devices moved.

    Open the original
  6. [20]

    MGM Resorts · Form 8-K

    5 October 2023. About US$100m impact on adjusted property EBITDAR.

    Open the original
  7. [24]

    Hong Kong Government · LCQ9: Combating frauds involving deepfake

    26 June 2024. A pre-recorded fake video conference and about HK$200m in losses.

    Open the original
  8. [25]

    CNN Business · Arup revealed as victim of US$25 million deepfake scam

    16 May 2024. Arup confirmed fake voices and images were used.

    Open the original
  9. [37]

    CISA, NSA and partners · Careful adoption of agentic AI services

    1 May 2026. Joint guidance from US, Australian, Canadian, New Zealand and UK agencies.

    Open the original
  10. [61]

    NIST · Cybersecurity Framework 2.0 and its quick-start guides, including SP 1300 for small businesses

    February 2024. Six functions: Govern, Identify, Protect, Detect, Respond, Recover. Govern was added in 2.0.

    Open the original
  11. [62]

    NIST · Cyber AI Profile (IR 8596, initial preliminary draft)

    16 December 2025. Three focus areas over CSF 2.0: secure AI systems, use AI to defend, and thwart AI-enabled attacks.

    Open the original
  12. [63]

    Center for Internet Security · CIS Controls v8.1 Implementation Groups

    2024. IG1: 56 safeguards against the most common attacks. IG2 adds 74; IG3 adds 23 more.

    Open the original
  13. [64]

    NIST · Zero Trust Architecture (SP 800-207)

    August 2020. Moves defenses from network perimeters to users, assets and resources.

    Open the original
  14. [65]

    OWASP GenAI Security Project · OWASP Top 10 for Agentic Applications for 2026

    9 December 2025. Ten risks, from agent goal hijack (ASI01) to rogue agents (ASI10).

    Open the original
  15. [66]

    Cyber Security Agency of Singapore · Securing Agentic AI: addendum to the Guidelines on Securing AI Systems

    22 October 2025. Released for consultation; maps agentic workflows to risks and controls across the lifecycle.

    Open the original
  16. [67]

    ISO · ISO/IEC 42001:2023, AI management system

    December 2023. A certifiable standard for managing AI in an organization.

    Open the original
  17. [68]

    MITRE · ATLAS: adversarial threat landscape for AI systems

    Ongoing. A knowledge base of attacker tactics and techniques against AI systems.

    Open the original
  18. [69]

    Meyer and coauthors, Microsoft · How effective is multifactor authentication at deterring cyberattacks?

    May 2023. Accounts showing suspicious activity: 99.22% lower risk of compromise, 98.56% where passwords had leaked.

    Open the original
  19. [70]

    UK National Cyber Security Centre · NCSC Annual Review 2025: empowering organisations

    2025. Insurer data on companies certified to Cyber Essentials; a correlation, not a controlled study.

    Open the original
  20. [71]

    Bono, Microsoft · Randomized controlled trials for a phishing triage agent

    November 2025. 167 external analysts; up to 6.5× true positives per analyst minute and 77% better F1 accuracy.

    Open the original
  21. [72]

    Chandra and coauthors · Deepfake-Eval-2024: an in-the-wild benchmark of deepfakes circulated in 2024

    March 2025, revised May 2026. Open-source detectors' AUC fell 50% (video), 48% (audio) and 45% (image) against earlier benchmarks.

    Open the original
  22. [73]

    Google Mandiant · M-Trends 2026: data, insights and strategies from the frontlines

    24 March 2026. Global median dwell time 14 days; exploits 32% of initial infections; 52% detected internally.

    Open the original