AI phishing & social engineering
High-quality personalised messages can be generated cheaply and at scale.
- Independent verification
- Phishing-resistant authentication
- Payment/change controls
Adversary AI / 2026
Assume attackers can move faster. Make trust harder to fake.
AI is lowering the cost of convincing content, analysis and automation. The answer is stronger identity, faster remediation, independent verification, useful telemetry and rehearsed containment.
IdentityVerify
ExposureReduce
TelemetryObserve
ResponseRehearse
Attack patterns
AI accelerates existing cyber activity and pressure.
Use controls that still stop the outcome when messages, voices, code and reconnaissance become more convincing.
Cloned voice & deepfake video
Familiar audio or video can create false confidence.
- Known callback channel
- Second approval
- Shared verification phrases
Automated reconnaissance
AI can summarise public and technical exposure faster.
- Reduce exposure
- Asset inventory
- Monitor leaked credentials
Credential attacks
AI can improve targeting around login and recovery workflows.
- Passkeys/phishing-resistant MFA
- Session monitoring
- Recovery hardening
Vulnerability discovery & chaining
Capable models can help find and combine weaknesses faster.
- Patch high-risk issues
- Harden configurations
- Reduce attack paths
Malicious-code assistance
AI can lower barriers for modifying or understanding code.
- Application control
- Endpoint detection
- Least privilege and segmentation
Defensive priorities
Shorten the distance between signal and safe action.
Combine prevention with fast verification and containment.
Identity first
Protect high-impact accounts and recovery paths.
Patch and harden
Reduce weaknesses that AI can discover but does not create.
Out-of-band verification
Move consequential requests to a separate trusted channel.
Useful detection
Prioritise signals that change a decision.
Rehearsed response
Know who can revoke access, isolate systems and preserve evidence.
Leadership questions
AI changes speed, scale and risk tolerance.
Leaders should test whether current assumptions hold when adversaries can analyse and act faster.
Exposure
Could AI-assisted attackers discover weaknesses faster than we remediate?
Chaining
Could several minor weaknesses combine into serious impact?
Supply chain
Do we understand AI-provider and technology dependencies?
Response speed
Can detection and incident decisions keep pace?
Exercises
Have response and continuity plans been tested against AI-enabled scenarios?
Human-controlled AI defence
Build capability without losing authority.
Start with the real workflow, data, identities, tools and consequences. Then place controls where they can stop harm.