NCSC warns AI to escalate cyber threats by 2027
NCSC assesses AI will make cyber intrusions more frequent and effective by 2027, widening the attack surface and deepening a security divide between defenders.
Sourcencsc.gov.uk — open the original
What to check
- Applies to
- Whether this concerns you is not assessed; the names the report mentions are listed below.
The UK's National Cyber Security Centre has published a forward-looking assessment on how artificial intelligence will shape the cyber threat landscape through 2027. The report, building on a January 2024 baseline assessment, concludes that AI will almost certainly make cyber intrusion operations more effective and efficient across reconnaissance, vulnerability research, exploit development, social engineering, malware generation, and data exfiltration—primarily by enhancing existing TTPs rather than creating novel attack vectors. NCSC judges that only the most capable state actors will fully harness AI for advanced operations in the near term, while most other threat groups will repurpose commercial and open-source AI models to uplift their capabilities, and a proliferation of AI-enabled tools will extend intrusion capability to a wider range of state and non-state actors, including cybercriminals offering AI-enabled tools 'as a service' to novices and hacktivists.
A key concern raised is AI-assisted vulnerability research and exploit development (VRED), which is expected to shrink the already narrow window between vulnerability disclosure and exploitation, increasing risk to unpatched systems and particularly to critical national infrastructure (CNI) and its supply chains, including operational technology with weaker security. The report also highlights the growing integration of AI systems into UK technology infrastructure as an expanding attack surface, citing threats such as direct and indirect prompt injection, software vulnerabilities, and supply chain attacks against AI systems themselves, compounded by insecure data handling, weak encryption, and poor identity management.
NCSC anticipates a growing 'digital divide' between organisations that keep pace with AI-enabled defensive capability and those that fall behind, warning that keeping pace with frontier AI will be critical to cyber resilience for the coming decade. While fully automated end-to-end attacks are considered unlikely by 2027, skilled actors are expected to increasingly automate elements of the attack chain—vulnerability exploitation, malware mutation, and infrastructure changes—to evade detection, a human-machine teaming trend that will complicate threat detection and mitigation absent comparable AI-enabled defensive tools.
Detection guidance
AI-Assisted Malware Generation via Command-Line Execution
A behavioural hunt for the technique in general.
Detects execution of AI model tools or Python scripts commonly used for malware obfuscation and polymorphic code generation, indicated by command-line patterns invoking AI frameworks with obfuscation-related arguments. Auto-generated starting point — validate and tune in your environment before deploying. IOC matches can false-positive on shared infrastructure and decay as adversary infrastructure rotates.
title: AI-Assisted Malware Generation via Command-Line Execution
description: Detects command-line execution of Python interpreters or AI model tools
with patterns indicative of malware obfuscation or polymorphic generation - such
as invoking LLM APIs, code mutation libraries, or obfuscation frameworks. Generalises
on the parent/child relation and suspicious argument patterns, not specific file
hashes or model names.
tags:
- attack.stealth
- attack.t1027
logsource:
category: process_creation
product: windows
detection:
selection_python_ai:
Image|endswith:
- \python.exe
- \python3.exe
CommandLine|contains:
- obfuscate
- mutate
- polymorphic
- llm
- gpt
- transform
- encode
selection_suspicious_tools:
Image|endswith:
- \ollama.exe
- \transformers
filter_legitimate_dev:
CommandLine|contains:
- --help
- version
- test
ParentImage|endswith:
- \git.exe
- \devenv.exe
condition: (selection_python_ai or selection_suspicious_tools) and not filter_legitimate_dev
falsepositives:
- Legitimate machine learning development environments using these libraries for benign
purposes
- Data science workflows involving code transformation or encoding
level: medium
id: 2ec017f1-e9ea-5ecb-a236-0851005c047a
status: experimental
author: Vorant
Behavioural rules are generated from public reporting — validate in your environment before deploying.
Source reporting: https://www.ncsc.gov.uk/report/impact-ai-cyber-threat-now-2027
Free public brief. The detections with Splunk SPL already written (plus Elastic and KQL where they apply) are in the app with a free account, no card. Slack alerts on the vendors you run come with the paid plans, and a new account gets three days of them.
What this brief leaves out
This page is the free tier: the write-up, the severity, the names. In the app the same report carries its detections (Splunk SPL already written, plus Elastic and Microsoft KQL where they apply). The paid plans add the security research and news reporting the free tier leaves out.
It sits in a corpus of 16,445 reports from 149 sources, 666 of them written in the last seven days, and it grows through the day.
A new account starts with three days of all of it, no card, then keeps the government and CERT reporting free. Slack alerts are a paid-plan feature: name the vendors you run and Slack hears about the reports that touch them, not the rest of the week's. Nothing is charged unless you subscribe.
Create a free account What it costs