AI is neither a doomsday nor a cure-all. It changes the speed and cost of attacks and gives defenders new leverage — while creating a fresh risk surface in your own use of it.
Day 3AI threatAI tool
Notes: Set a balanced tone from the start — neither hype nor dismissal. Three lenses: threat, tool, and your own risk surface.
Lesson 11 · Slide 2
What AI actually changes
It rarely invents wholly new attacks
It makes existing attacks faster, cheaper, and more convincing
It lowers the skill needed to attempt them
So the fundamentals (Lessons 04, 09) still hold — they just face more, better attempts
The reframe
AI shifts the economics of attacks. Your defences don't become obsolete; they get tested more often and more convincingly.
Notes: This is the core executive insight and prevents panic. Reveal each.
Lesson 11 · Slide 3
AI as threat
Flawless social engineering — localised, personalised phishing at scale
Real-time deepfakes — voice and video clones (recall Lesson 09)
Lowered barrier — less-skilled actors attempt more capable attacks
Notes: Reveal each. Note the callbacks — AI supercharges threats we've already studied. Keep malware discussion conceptual, no code.
Lesson 11 · Slide 4
The AI-accelerated attack chain
Same chain as ever — just faster and more convincing at every step. Press →.
AI aggregates your public footprint in minutes
→
Tailored deepfake / flawless phish
→
Faster exploitation & iteration
Notes: Reveal step by step. The hot node — the tailored deepfake/phish — is where AI's advantage is sharpest and where verification (Lesson 09) is the answer.
Lesson 11 · Slide 5
AI as tool for defenders
Detection & triage — surface the signal in a flood of alerts
Summarisation — turn noisy data into decisions faster
Response & hunting — accelerate investigation and containment
Augments, never replaces — human judgment stays in the loop
The balance
AI is a force multiplier for defenders too. The winning posture uses it — with humans owning the consequential calls.
Notes: Balance the threat slide. The caveat (augment, not replace) is essential and sets up governance.
Lesson 11 · Slide 6
AI as your risk surface
Data leakage — staff pasting secrets into public AI tools
Prompt injection — hostile input hijacking an AI's instructions
Output trust — confident, plausible, and sometimes wrong
Shadow AI — tools adopted with no oversight
The quiet risk
Your biggest near-term AI exposure may not be the attacker's AI — it may be your own people feeding sensitive data into tools you don't govern.
Notes: This is the under-appreciated third lens. Sets up the poll. Data leakage + deepfakes are most orgs' realistic top risks.
Lesson 11 · Slide 7
Which is the biggest near-term AI risk to most organisations?
C. A and B exist but are still specialised and rarer; D is science fiction, not a security plan. For most organisations today, the realistic damage comes from convincing deepfake/social-engineering fraud (defended by Lesson 09's verification rule) and from employees pasting confidential data into ungoverned AI tools. Govern the mundane risks first.
Notes: Push the room past the sci-fi answers. The defensible position: deepfakes + data leakage are the here-and-now, and both are governable.
Lesson 11 · Slide 8
Governing AI use
Acceptable-use rules — what may and may not go into which tools
Data classification first — know what's sensitive before it's typed into a prompt
Human-in-the-loop — a person owns every consequential decision
Recognised frameworks — e.g., NIST AI RMF — to structure the approach
Notes: Governance is the leader's lever. Reveal each. Frameworks referenced as concepts, not products.
Lesson 11 · Slide 9
Deepfakes: fundamentals still win
AI makes the clone perfect, but the defence never relied on spotting fakes. The Lesson 09 rule holds: verify the request on an independent channel with a pre-agreed secret.
The reassurance
Better fakes do not defeat process-based verification. They only defeat "I trusted my ears."
Notes: Reinforce continuity with Day 2. AI raises the stakes but not the solution — verification is unchanged.
Lesson 11 · Slide 10
Two-minute drill: your acceptable-AI-use line
Write one sentence of your team's acceptable-AI-use rule — what kind of information must never be pasted into a public AI tool.
Notes: Start the timer. A single clear line ("never paste client, personal, or classified data into public tools") prevents most leakage.
Lesson 11 · Slide 11
Recap & what's next
AI changes the economics of attacks; fundamentals still defend you
Use AI as a defender — with humans owning consequential decisions
Govern your own AI use: acceptable-use, data classification, human-in-the-loop
Next
Lesson 12 — Ransomware & Destructive Malware: the crisis that tests every foundation at once.
Notes: Bridge: AI accelerates the delivery of exactly the kind of intrusion that ends in ransomware — the next lesson.