AI-made malware: hype and reality
AI can help attackers write code and lures faster. But most attacks still use known techniques, and the basic defences still work.
In 60 seconds
AI-made malware: hype and reality
AI can help attackers write code and lures faster. But most attacks still use known techniques, and the basic defences still work.
What AI realistically helps attackers do
- Write or fix code faster, like it helps any programmer.
- Write convincing lures: the email, fake document or fake website that tricks someone into running the malware.
- Change the look of existing malware so simple signature checks miss it.
- Read and summarise stolen data quickly to find what's valuable.
- Research targets, such as which software a company uses.
What mostly stays the same
- AI malware is unstoppable
- Only AI security tools can help now
- Basic hygiene no longer matters
- Every attack is now new and unknown
- It still needs a way in and still leaves traces
- AI tools can help, and also make mistakes and can be attacked
- Patching, MFA and backups block many attacks
- Most attacks reuse well-known techniques
What defenders should keep doing
| Control | Why it still works |
|---|---|
| Patching and updates | Closes the known holes attackers most often use. |
| MFA, ideally phishing-resistant | A stolen password alone is not enough to log in. |
| Backups, offline or protected | You can recover from ransomware without paying. Test restores regularly. |
| Endpoint detection and response (EDR) | Watches behaviour on laptops and servers, not just known file signatures, so changed malware can still be caught. |
| Least privilege | Users and apps get only the access they need, so malware can reach less. |
| Limit macros and unknown software | Blocks common ways malware starts on a laptop. |
| Asset list | You can't protect devices and accounts you don't know you have. |
A simple check you can run
Get-HotFix | Sort-Object InstalledOn -Descending | Select-Object -First 10Watch and read more
Lab
A one-page 'basics scorecard' for a small organisation, with your top 3 fixes.
The problem
Use your own organisation, a small business you know, or a made-up 20-person office. Work on paper; do not scan or test any system you don't own and manage.
- Make a table with these rows: patching, MFA, backups (and restore tests), EDR or antivirus, admin rights, macros and unknown software, asset list.
- For each row, score 0 (not in place), 1 (partly) or 2 (fully), with one line of evidence.
- Pick the 3 lowest-scoring items and write a first step for each that could be done this month.
- Write 2 sentences explaining to a manager why these matter more than buying a new "AI security" tool.
You are done when
Hard questions
Try to answer before you reveal. If you can answer these, you understood the lesson.
Q1If AI can quickly change how malware looks, why does behaviour-based detection still help?Reveal
Changing the file's appearance doesn't change what it has to do: run, connect out, gain access, encrypt or copy data. Those actions leave patterns that behaviour-based tools and logs can notice. Least privilege and backups also work regardless of what the malware looks like.
Q2An executive says 'AI attacks are so advanced that patching is pointless now.' How do you respond?Reveal
Explain that AI mostly helps attackers move faster through the same doors: phished passwords, missing MFA, unpatched systems. Closing those doors is still the most effective, affordable step. Faster attackers make patching and MFA more urgent, not less. New tools can be added on top, but not instead.
Questions people ask
Should we buy an AI security product to fight AI malware?
AI-based tools can help, especially with detecting unusual behaviour. But they are not a replacement for basics, and they can make mistakes. Ask vendors for evidence, try them on your own environment, and fix gaps in patching, MFA and backups first.
Can antivirus still catch AI-changed malware?
Simple signature checks may miss a changed file. Modern endpoint tools also look at behaviour, such as a program suddenly encrypting many files, which is harder to hide. That's why EDR and least privilege matter.
What is MITRE ATT&CK?
It's a public knowledge base of techniques attackers use, grouped by stage (getting in, staying, moving around, stealing data). Defenders use it to check which techniques they can detect or block. You can read it at attack.mitre.org.
We had a ransomware alert. What do we do?
Disconnect affected devices from the network (don't switch them off unless your IT team says so), tell your IT or security team and management, and follow your incident plan. Organisations in India may also need to report certain cyber incidents to CERT-In quickly; check CERT-In's current directions with your IT or legal team.
Lesson test
5 questions. Get 3 right (60%) to pass and complete this lesson.