Deepfakes, Detection & Content Authenticity

The fraud patterns that actually happen, why detectors fail, out-of-band verification, provenance standards, and what to do if you are targeted.

The threat is not a perfect Hollywood forgery. It is a 30-second voice note that is good enough in a moment of urgency, which is why this course focuses on process rather than analysis. You will learn the six fraud patterns that actually occur, from the executive payment request to the fake celebrity endorsement that drove India's February 2026 deepfake framework, and the structure they all share. Then the uncomfortable truth: AI detection tools are unreliable in both directions and must never be the basis of an accusation. What works is out-of-band verification, pre-agreed challenge phrases and dual authorisation with no urgency exception. It covers content provenance honestly, including why absence of metadata proves nothing, and what to do if synthetic content targets you.

What you'll learn

• Calibrate what synthetic media can and cannot currently do
• Recognise the six real fraud patterns by their shared structure
• Verify out of band instead of relying on unreliable detectors
• Use and preserve content provenance, and know its limits
• Respond correctly if you or your organisation is targeted
• Hold calibrated scepticism without falling for the liar's dividend

Course content

1. 1. What synthetic media can now do (15 min)
2. 2. The fraud patterns that actually happen (15 min)
3. 3. Verification that works, and detection that does not (15 min)
4. 4. Content authenticity and provenance (15 min)
5. 5. If it happens to you or your organisation (15 min)
6. 6. Living with an unreliable information environment (15 min)

Related courses

• AI Ethics & Responsible Use at Work — Four questions that resolve real workplace decisions, plus bias, confidentiality, disclosure, and a one-page policy people will actually follow.
• AI Compliance Essentials: EU AI Act, DPDP & the Literacy Obligation — Enforcement began August 2026. Risk tiers, what Article 4 literacy actually requires, transparency duties, India's DPDP and deepfake framework, and a runnable programme.
• AI Red-Teaming & Model Safety Evaluation for Language Experts — Paid evaluation work on Outlier, DataAnnotation and Alignerr: preference ranking, rubric consistency, red-team reporting, and the multilingual gaps labs cannot test without you.